https://reading-list.oddship.net A curated linklog of essays, posts, papers, and notes. Wed, 26 Aug 2026 21:45:00 +0530 https://reading-list.oddship.net/notes/2026-08-26-ducklabs-joins-aws/ https://reading-list.oddship.net/notes/2026-08-26-ducklabs-joins-aws/ Wed, 26 Aug 2026 21:45:00 +0530 Logged at IST: 2026-08-26 21:45 IST What it is: DuckLabs' announcement that it is joining AWS, shared by Frank Denis. Gist: DuckLabs says the Amsterdam team behind DuckDB will stay together under AWS, continuing work on DuckDB, DuckLake, Quack, and the wider Duck Stack. The company is presenting this as a scale move: DuckDB now has more than one million downloads a day, and DuckLabs says AWS gives it the resources, infrastructure, and customer reach to grow without turning the small team into a large sales and support organization. The trust question is the important part. DuckLabs says DuckDB… https://reading-list.oddship.net/notes/2026-08-26-go-sync-map-hash-trie/ https://reading-list.oddship.net/notes/2026-08-26-go-sync-map-hash-trie/ Wed, 26 Aug 2026 16:33:00 +0530 Logged at IST: 2026-08-26 16:33 IST What it is: Phuong Le's VictoriaMetrics walkthrough of Go's hash-trie-backed sync.Map implementation. Gist: The visible API is still the familiar Load, Store, Delete, LoadOrStore, CompareAndSwap, Range, and Clear. Underneath, Go moved sync.Map to an internal HashTrieMap: keys are hashed, the hash is consumed in 4-bit chunks, each indirect node has 16 child slots, and full hash collisions fall back to overflow entries. The concurrency model is the useful part. Load can walk atomic child pointers without taking node mutexes. Writers search without a lock, then… https://reading-list.oddship.net/notes/2026-08-26-go-generic-methods-interface-boundary/ https://reading-list.oddship.net/notes/2026-08-26-go-generic-methods-interface-boundary/ Wed, 26 Aug 2026 10:14:00 +0530 Logged at IST: 2026-08-26 10:14 IST What it is: Dominik Honnef's explanation of generic methods in Go 1.27. Gist: Go 1.27 removes a long-standing awkwardness in Go generics: methods on concrete types can now declare their own type parameters. That means an API like Node[T].Map[U] no longer has to pollute the receiver type with a U that only one method needs, or move the operation out into a package-level helper function. The interesting part is the boundary Go keeps. Interfaces still cannot declare generic methods, and a concrete generic method does not satisfy a matching interface method. Hon… https://reading-list.oddship.net/notes/2026-08-25-go-json-v2-migration-guide/ https://reading-list.oddship.net/notes/2026-08-25-go-json-v2-migration-guide/ Tue, 25 Aug 2026 20:19:00 +0530 Logged at IST: 2026-08-25 20:19 IST What it is: Anton Zhiyanov's interactive guide to Go's new json/v2 package. Gist: The guide is a migration map from encoding/json v1 to json/v2. The basics still look familiar, but the surrounding API changes a lot: direct MarshalWrite and UnmarshalRead, streaming encode/decode through jsontext, composable options instead of many one-off functions, new struct tags like case and embed, and streaming custom marshaling through MarshalJSONTo / UnmarshalJSONFrom. The defaults also change in ways that will matter in real services. Nil slices encode as [] instead o… https://reading-list.oddship.net/notes/2026-08-25-go-llama-pure-go-local-inference/ https://reading-list.oddship.net/notes/2026-08-25-go-llama-pure-go-local-inference/ Tue, 25 Aug 2026 20:15:00 +0530 Logged at IST: 2026-08-25 20:15 IST What it is: goccy/go-llama, a pure-Go llama.cpp-style inference library for running GGUF models from Go applications. Gist: The interesting bit is the packaging path. go-llama does not bind to llama.cpp through cgo, and it does not embed a wasm runtime. Instead, llama.cpp is compiled to WASI WebAssembly, then translated ahead-of-time into standalone Go through wasm2go and llamawasm2go. The result is meant to be a normal Go dependency: no shared library, no cgo, and static-binary friendly. The README claims a fairly complete local-inference surface: model/con… https://reading-list.oddship.net/notes/2026-08-25-private-agent-dev-repo/ https://reading-list.oddship.net/notes/2026-08-25-private-agent-dev-repo/ Tue, 25 Aug 2026 12:35:00 +0530 Logged at IST: 2026-08-25 12:35 IST What it is: Ryan Dahl describing how celld is developed from a larger private denoland/celld.dev repo, while the public denoland/celld repo stays a lean export. Gist: The useful part is the shape of the working repo. Dahl says the private repo has an LLM-maintained design wiki, a large test suite, a TLA+ spec, website code, cloud experiment scripts, and a speculative managed control plane. The public repo is open source, but the real development loop happens around the richer private artifact. That feels like an agent-era pattern: code is no longer the only … https://reading-list.oddship.net/notes/2026-08-24-yegge-fences-not-sandboxes/ https://reading-list.oddship.net/notes/2026-08-24-yegge-fences-not-sandboxes/ Mon, 24 Aug 2026 23:01:00 +0530 Logged at IST: 2026-08-24 23:01 IST What it is: Steve Yegge's essay on what his Wheelhouse agent organization for Wyvern has taught him about governing fleets of AI workers. Gist: Yegge's claim is that once Fable-class models get cheap enough for every company to run hundreds or thousands of AI employees, the control problem stops being mostly about tight sandboxes and narrow task wrappers. In his account, his agents gradually built something closer to a legal system: offices, roles, jurisdictions, rulings, case law, runbooks, gates, tripwires, authority envelopes, and mechanical checks that d… https://reading-list.oddship.net/notes/2026-08-23-how-complex-systems-fail/ https://reading-list.oddship.net/notes/2026-08-23-how-complex-systems-fail/ Sun, 23 Aug 2026 11:26:00 +0530 Logged at IST: 2026-08-23 11:26 IST What it is: Richard I. Cook's classic short treatise on the nature of failure in complex systems, shared by Jessica Paquette. Gist: Cook's frame is still the cleanest antidote to simple root-cause stories. Complex systems are intrinsically hazardous, but they are also heavily defended, so major failures usually require several small, individually insufficient faults to line up. The system may look normal because it runs in degraded mode all the time, with people constantly adapting around latent problems. That makes post-accident blame misleading. Hindsight … https://reading-list.oddship.net/notes/2026-08-23-fast-hard-code/ https://reading-list.oddship.net/notes/2026-08-23-fast-hard-code/ Sun, 23 Aug 2026 11:13:00 +0530 Logged at IST: 2026-08-23 11:13 IST What it is: Armin Ronacher's short post on how LLMs change language and technology choices when starting new projects. Gist: Ronacher's point is not that programming is solved. It is that the cost of becoming productive in an unfamiliar language has dropped enough that old language-choice friction matters less. If an agent can draft, translate, and optimize code in a language I barely know, I can choose more for project identity, performance goals, and the promise of small, fast software. That changes which technologies feel approachable. Rust and Zig benefi… https://reading-list.oddship.net/notes/2026-08-20-typescript-go-content-mappers/ https://reading-list.oddship.net/notes/2026-08-20-typescript-go-content-mappers/ Thu, 20 Aug 2026 18:37:00 +0530 Logged at IST: 2026-08-20 18:37 IST What it is: Microsoft’s typescript-go PR #4712, “Content mappers”, merged by Andrew Branch and called out as likely to land in TypeScript 7.1. Gist: Content mappers let TypeScript include otherwise unsupported file types in a program. A mapper turns a foreign file such as .vue, .svelte, .astro, or an Angular-style source into valid JS, JSX, TS, TSX, or JSON, then returns span mappings back to the original content. The PR defines the operational contract around that idea. Projects can declare mapper packages in tsconfig.json, mapper packages describe their ex… https://reading-list.oddship.net/notes/2026-08-20-sqlite-in-nix-wasm/ https://reading-list.oddship.net/notes/2026-08-20-sqlite-in-nix-wasm/ Thu, 20 Aug 2026 11:17:00 +0530 Logged at IST: 2026-08-20 11:17 IST What it is: Farid Zakaria's write-up on trying to expose SQLite-backed queries to Nix for nixpkgs-multiverse. Gist: nixpkgs-multiverse currently carries JSON indexes, but builtins.fromJSON is eager: one lookup pays the cost of parsing and materializing the whole file. Zakaria uses that as the setup for three SQLite-in-Nix experiments. builtins.exec is the simple unsafe escape hatch: ask sqlite3 to print a Nix expression and let the evaluator parse it. builtins.importNative goes further by loading a shared object that can cache SQLite handles and build Nix va… https://reading-list.oddship.net/notes/2026-08-20-go-1-27-release/ https://reading-list.oddship.net/notes/2026-08-20-go-1-27-release/ Thu, 20 Aug 2026 06:31:00 +0530 Logged at IST: 2026-08-20 06:31 IST What it is: The official Go 1.27 release post, shared through mattn's Japanese X thread of highlights. Gist: Go 1.27 is a practical release across the language, toolchain, runtime, and standard library. The language headline is generic methods, alongside broader generic function inference and direct promoted-field keys in struct literals. The operational pieces are just as interesting. go fix gets more modernizers, go doc can query package@version, go mod tidy normalizes require blocks, small object allocation gets faster, and the goroutine leak profile is n… https://reading-list.oddship.net/notes/2026-08-19-rangebitmap-slicez-kubernetes-metadata/ https://reading-list.oddship.net/notes/2026-08-19-rangebitmap-slicez-kubernetes-metadata/ Wed, 19 Aug 2026 14:28:00 +0530 Logged at IST: 2026-08-19 14:28 IST What it is: Richard Startin points to a Yandex Monium write-up using RangeBitmap, a range-encoded bit-sliced index he designed for Apache Pinot and which is available in RoaringBitmap. His follow-up links SliceZ, his newer Java successor for range and point queries over unsorted numeric data. Gist: Yandex Monium stores billions of metrics, processes about 50 GB/s of logs, and serves around 16,000 internal users. Kubernetes broke an old metadata assumption: pod labels are short-lived, but historical label values accumulate. After a few weeks, selectors can hi… https://reading-list.oddship.net/notes/2026-08-19-git-at-any-scale-cursor/ https://reading-list.oddship.net/notes/2026-08-19-git-at-any-scale-cursor/ Wed, 19 Aug 2026 09:51:00 +0530 Logged at IST: 2026-08-19 09:51 IST What it is: Vicent Martí's Cursor post on the storage system behind Origin, Cursor's Git hosting product. Gist: The useful framing is that Git hosting is hard because Git was designed so every repository copy is equivalent. That makes the local developer workflow great, but it makes server-side scaling awkward: packfiles are optimized for local disk and Git clients still expect packfiles over the network. The post walks through prior approaches. Object-level distributed stores map nicely to content-addressed Git objects, but DAG walks become round-trip-heavy… https://reading-list.oddship.net/notes/2026-08-19-postgres-poisoned-connection-pools/ https://reading-list.oddship.net/notes/2026-08-19-postgres-poisoned-connection-pools/ Wed, 19 Aug 2026 09:33:00 +0530 Logged at IST: 2026-08-19 09:33 IST What it is: PlanetScale's debugging note on poisoned Postgres connection pools, shared by Ben Dicken. Gist: The failure mode is simple and nasty: PgBouncer transaction pooling reuses underlying database connections across clients. If one client leaves session state behind, the next client can inherit it. The concrete example is read-only state. Application code used SET SESSION CHARACTERISTICS AS TRANSACTION READ ONLY for a read-only transaction, but that changed the session rather than only the transaction. Later requests reused that same underlying connect… https://reading-list.oddship.net/notes/2026-08-19-github-outage-rca-retry-storm/ https://reading-list.oddship.net/notes/2026-08-19-github-outage-rca-retry-storm/ Wed, 19 Aug 2026 09:31:00 +0530 Logged at IST: 2026-08-19 09:31 IST What it is: GitHub's RCA for the August 17, 2026 critical GitHub.com incident, shared by Cassidy Williams. Gist: GitHub.com saw elevated errors and latency for 7h47m across Issues, Pull Requests, APIs, Actions, and Copilot. At peak, web and API error rates were around 20%, while archive and raw-content downloads reached around 50%. The immediate cause was load-balancer saturation in Central US. An Istio sidecar pod hit its concurrency limits and did not autoscale correctly because the policy watched host service capacity but not sidecar capacity. That cascad… https://reading-list.oddship.net/notes/2026-08-18-ai-plays-age-of-empires-ii/ https://reading-list.oddship.net/notes/2026-08-18-ai-plays-age-of-empires-ii/ Tue, 18 Aug 2026 16:47:00 +0530 Logged at IST: 2026-08-18 16:47 IST What it is: Emergent Garden's video on using modern AI agents to play Age of Empires II, along with the linked AgentsOfEmpires runner repo. Gist: The video description says the creator used LLM-powered agents, including Claude, GPT, Gemini, and Kimi K3, to write custom Age of Empires II AI scripts, make them fight, and then tune strategies over repeated runs. The listed chapters move from "Agents of Empires" into playing the game, model comparisons, strategy optimization, and battle optimization. The linked GitHub repo is useful context because it shows the … https://reading-list.oddship.net/notes/2026-08-18-detecting-goroutine-leaks/ https://reading-list.oddship.net/notes/2026-08-18-detecting-goroutine-leaks/ Tue, 18 Aug 2026 14:15:00 +0530 Logged at IST: 2026-08-18 14:15 IST What it is: Anton Zhiyanov's walkthrough of goroutine leak detection in modern Go, comparing goleak, synctest, and the experimental goroutineleak pprof profile. Gist: The article starts with the practical shape of a goroutine leak: goroutines blocked forever on synchronization primitives while the rest of the program keeps running. That makes leaks quieter than deadlocks and harder to catch than data races. Zhiyanov walks through a simple unbuffered-channel example, then compares three ways to find the leak. goleak works today as a third-party test helper. s… https://reading-list.oddship.net/notes/2026-08-18-video-based-cryptanalysis/ https://reading-list.oddship.net/notes/2026-08-18-video-based-cryptanalysis/ Tue, 18 Aug 2026 14:11:00 +0530 Logged at IST: 2026-08-18 14:11 IST What it is: Ben Nassi, Etay Iluz, Ofek Vayner, Or Cohen, Dudi Nassi, Boris Zadov, and Yuval Elovici's 2023 work on recovering cryptographic keys from video footage of a device's power LED. Gist: The paper's useful idea is that a status LED can carry more information than it appears to. Cryptographic work changes a device's power consumption. In many circuits, the power LED's brightness or color shifts with that consumption, so video of the LED can become a side-channel trace. The researchers show how to turn ordinary video into a higher-rate signal by fillin… https://reading-list.oddship.net/notes/2026-08-17-software-cartographers/ https://reading-list.oddship.net/notes/2026-08-17-software-cartographers/ Mon, 17 Aug 2026 11:37:00 +0530 Logged at IST: 2026-08-17 11:37 IST What it is: Adam Craven's 2023 essay arguing for "software cartographers": people and processes that keep maps of working software in sync with the code. Gist: The essay uses the village, town, and city analogy to explain why small codebases fit inside one engineer's head while large systems stop being knowable by any individual, or eventually by the whole team. Other complex human-built systems use maps to reduce complexity. Software usually has diagrams at the start, then lets them rot once code becomes the only source of truth. The practical claim is that… https://reading-list.oddship.net/notes/2026-08-14-miren-rpc-that-calls-you-back/ https://reading-list.oddship.net/notes/2026-08-14-miren-rpc-that-calls-you-back/ Fri, 14 Aug 2026 10:05:00 +0530 Logged at IST: 2026-08-14 10:05 IST What it is: Miren engineering post by Evan Phoenix and Paul Hinze on the RPC layer they built for typed two-way calls across CLI, control plane, and runtime components. Gist: Miren argues that a lot of deployment-platform work does not fit cleanly into one request and one response. A deploy has progress updates, health checks, subscriptions, and callbacks from a server to an ephemeral client. Polling, webhooks, or a separate stream work, but they make the request/response model leak. Their answer is object-capability RPC. In the IDL, parameters and results c… https://reading-list.oddship.net/notes/2026-08-13-go-sync-nocopy/ https://reading-list.oddship.net/notes/2026-08-13-go-sync-nocopy/ Thu, 13 Aug 2026 17:56:00 +0530 Logged at IST: 2026-08-13 17:56 IST What it is: func25.dev explaining Go's internal sync.noCopy marker and how go vet detects accidental copies of sync-like structs. Gist: noCopy is not compiler magic. It is an empty marker with two empty methods on its pointer receiver: Lock and Unlock. Those names make *noCopy satisfy sync.Locker, while the value noCopy does not. go vet's copylocks checker uses that shape: while inspecting copied types, it recursively looks for a type whose pointer implements sync.Locker but whose value does not. That is why go build will still accept copying a used sync.Map… https://reading-list.oddship.net/notes/2026-08-13-appwrite-cli-go-rewrite/ https://reading-list.oddship.net/notes/2026-08-13-appwrite-cli-go-rewrite/ Thu, 13 Aug 2026 09:59:00 +0530 Logged at IST: 2026-08-13 09:59 IST What it is: Richard Seroter pointing to Appwrite's rewrite of its CLI from TypeScript/Bun to Go. Gist: Appwrite says the user-facing contract stays the same: npm package name, appwrite binary, flags, scripts, CI behavior, exit codes, and JSON output. The internals changed from a JavaScript bundle on a runtime plus 189 packages underneath to a native Go binary. The measured deltas are the useful part: startup went from 207.6ms to 11.1ms, install dependency count from 330 packages to 2, on-disk install from 209MB to 13MB, memory during push from 283.5MB to 28.… https://reading-list.oddship.net/notes/2026-08-13-tailscale-sqlite-wal-reset-bug/ https://reading-list.oddship.net/notes/2026-08-13-tailscale-sqlite-wal-reset-bug/ Thu, 13 Aug 2026 08:09:00 +0530 Logged at IST: 2026-08-13 08:09 IST What it is: Tailscale’s postmortem on months of control-plane instability caused by a rare SQLite WAL checkpoint race, and the follow-on false alarm from stale expression indexes. Gist: Tailscale runs its control plane as isolated shards, each backed by a single-writer SQLite database. That architecture was supposed to be boring, but their backup pipeline started finding corrupted SQLite files. Over six months they saw 19 corruption incidents, causing shard-local control-plane downtime and occasional loss of recent configuration metadata. The bug had no easy… https://reading-list.oddship.net/notes/2026-08-12-html-over-wire-sse-websockets/ https://reading-list.oddship.net/notes/2026-08-12-html-over-wire-sse-websockets/ Wed, 12 Aug 2026 19:04:00 +0530 Logged at IST: 2026-08-12 19:04 IST What it is: A YAGNI Club response to an “HTML over WebSockets” article, arguing that many HTML-over-the-wire apps should consider SSE plus normal POSTs before reaching for WebSockets. Gist: The core correction is transport-level. The original WebSocket argument says one persistent connection avoids TCP and HTTP overhead, but with HTTP/2 keep-alive the page, CSS, SSE stream, and POST commands can already share the same connection. A WebSocket often means a second TCP/TLS connection plus an HTTP/1.1 upgrade, while command POSTs keep the normal HTTP request lif… https://reading-list.oddship.net/notes/2026-08-12-compression-is-prediction/ https://reading-list.oddship.net/notes/2026-08-12-compression-is-prediction/ Wed, 12 Aug 2026 19:00:00 +0530 Logged at IST: 2026-08-12 19:00 IST What it is: Armin Ronacher recommending Annie Sexton’s ngrok explainer, “Compression is prediction.” Gist: The piece explains compression through the same lens as language modeling: a model predicts symbol probabilities, and an entropy coder turns those probabilities into a bitstream. Arithmetic coding makes the connection especially clear: high-probability symbols keep the encoded range wider and cost fewer bits; low-probability misses shrink the range and require more precision. The useful bridge to LLMs is that language models are also next-token probabil… https://reading-list.oddship.net/notes/2026-08-12-hcl-founding-story-ajai-chowdhry/ https://reading-list.oddship.net/notes/2026-08-12-hcl-founding-story-ajai-chowdhry/ Wed, 12 Aug 2026 10:13:00 +0530 Logged at IST: 2026-08-12 10:13 IST What it is: Ajai Chowdhry, HCL cofounder, recounting how the original six founders left their jobs in 1976 to start Microcomp, which became Hindustan Computers Limited through a licence partnership with UP Electronics Corporation. Gist: The useful part is the pre-venture-capital, pre-mainstream-computing texture: six founders pooled ₹1.87 lakh, started from a barsati in Golf Links, and tried to build an Indian computer around the microprocessor before most customers knew what that meant. The licence regime forced the UP Electronics partnership, and the go-to… https://reading-list.oddship.net/notes/2026-08-12-nvidia-risky-business-ai-infrastructure-financing/ https://reading-list.oddship.net/notes/2026-08-12-nvidia-risky-business-ai-infrastructure-financing/ Wed, 12 Aug 2026 00:12:00 +0530 Logged at IST: 2026-08-12 00:12 IST What it is: Mario Zechner recommending Ben Thompson’s Stratechery essay on Nvidia and AI infrastructure financing. Gist: Thompson frames the current AI buildout through the 1870s railroad-financing boom and Jay Cooke’s role in spreading Northern Pacific risk to retail investors. The analogy is not that AI equals railroads, but that huge capital needs create pressure to invent new funding mechanisms when ordinary cash flow and debt markets are not enough. The essay argues that hyperscalers have already moved beyond free-cash-flow-funded CapEx into large debt … https://reading-list.oddship.net/notes/2026-08-11-hetzner-experiments-open-weight-inference/ https://reading-list.oddship.net/notes/2026-08-11-hetzner-experiments-open-weight-inference/ Tue, 11 Aug 2026 20:03:00 +0530 Logged at IST: 2026-08-11 20:03 IST What it is: Hetzner's exploratory Experiments platform with an experimental open-weight LLM inference API. Gist: Hetzner says the API is free for now, has no SLAs, and is meant for users to test their own workloads and report what works. They are explicit that this may not become a permanent product. The linked site resolves to experiments.hetzner.com; the app exposes an AI Inference · Hetzner Experiments area, API-token creation, docs links, and an inference endpoint at https://inference.hetzner.com/api/v1. The unauthenticated /models endpoint returns 401 u… https://reading-list.oddship.net/notes/2026-08-10-source-code-availability/ https://reading-list.oddship.net/notes/2026-08-10-source-code-availability/ Mon, 10 Aug 2026 16:19:00 +0530 Logged at IST: 2026-08-10 16:19 IST What it is: Loris Cro’s long essay on source-code availability, using Zine, Zig, Radicle, Codeberg, Tangled, community mirrors, and package registries as examples. Gist: Cro starts from a practical build problem: if a project’s dependencies live across GitHub, Codeberg, and small self-hosted forges, fresh builds can fail whenever any host is down. Forking and vendoring solve this locally, but they do not make the copied code discoverable or automatically useful as shared redundancy. The core argument is that source availability is a cost-allocation problem. … https://reading-list.oddship.net/notes/2026-08-08-incus-agent-sandboxes/ https://reading-list.oddship.net/notes/2026-08-08-incus-agent-sandboxes/ Sat, 08 Aug 2026 19:05:00 +0530 Logged at IST: 2026-08-08 19:05 IST What it is: Incus is the Linux Containers project's system container, application container, and VM manager. Shantanu Goel's linked post applies it to AI agent sandboxes through a small incus-manager setup repo. Gist: Incus gives a public-cloud-like interface for running system containers, application containers, and virtual machines on shared storage and networking. It was created as a community-driven alternative to Canonical's LXD and is maintained by many of the same people who created LXD. For coding agents, the interesting mode is the system container:… https://reading-list.oddship.net/notes/2026-08-08-logchef-2-open-source-log-analytics/ https://reading-list.oddship.net/notes/2026-08-08-logchef-2-open-source-log-analytics/ Sat, 08 Aug 2026 19:00:00 +0530 Logged at IST: 2026-08-08 19:00 IST What it is: Karan Sharma's launch thread for Logchef 2.0, an open-source, self-hosted log analytics workspace for ClickHouse and VictoriaLogs. Gist: Logchef 2.0 expands the project from a fast ClickHouse log explorer into a broader operational workspace. The new shape is search, live tail, dashboards, alerts, RBAC, CLI, AI query assistant, and MCP integration in one self-hosted binary. The product choice I like here is that Logchef does not try to become another storage layer. It queries ClickHouse and VictoriaLogs directly, so existing ingestion, retention,… https://reading-list.oddship.net/notes/2026-08-08-metabase-critical-security-update/ https://reading-list.oddship.net/notes/2026-08-08-metabase-critical-security-update/ Sat, 08 Aug 2026 00:24:00 +0530 Logged at IST: 2026-08-08 00:24 IST What it is: Metabase’s security update for a critical actively exploited 0-day affecting Metabase versions 1.58 and above. Gist: Metabase says its Cloud service was attacked through an unknown security vulnerability in versions 1.58 and above. They blocked the endpoints used for the attack, identified the issue, and shipped patches. Metabase Cloud customers are already upgraded, but self-hosted instances may still be vulnerable. The linked advisory describes the bug as unauthenticated arbitrary SQL injection against the Metabase application database. That is… https://reading-list.oddship.net/notes/2026-08-07-database-backpressure-concurrency-throughput/ https://reading-list.oddship.net/notes/2026-08-07-database-backpressure-concurrency-throughput/ Fri, 07 Aug 2026 21:46:00 +0530 Logged at IST: 2026-08-07 21:46 IST What it is: Sam Lambert’s X post recommending Liz van Dijk’s PlanetScale engineering post on concurrency, throughput, Vitess, and MySQL. Gist: PlanetScale describes a production MySQL database that melted down for sixteen minutes after a batch job opened a transaction against a hot table, took row locks, and held them without committing. The obvious story would be “lock contention,” but the useful point is subtler. Many reads were not waiting on the locked rows. They were doing snapshot reads through an increasingly long version history, blowing past executi… https://reading-list.oddship.net/notes/2026-08-07-exe-dev-software-factory-inventory/ https://reading-list.oddship.net/notes/2026-08-07-exe-dev-software-factory-inventory/ Fri, 07 Aug 2026 16:17:00 +0530 Logged at IST: 2026-08-07 16:17 IST What it is: exe.dev’s X post pointing to Philip Zeyliger’s inventory of the internal agents, bots, and operational tools they use to run exe. Gist: The post is useful because it lists the actual places where exe.dev has put agents and small internal systems into its operating loop. They have agents for systematic security review, alert investigation, daily log-trend emails, and deploy supervision. They also have bots that look for flaky or slow tests, a homegrown status page, Pushover-based phone paging, and daily Slack reports about git commits and support/… https://reading-list.oddship.net/notes/2026-08-07-end-of-no-code-agentic-linux/ https://reading-list.oddship.net/notes/2026-08-07-end-of-no-code-agentic-linux/ Fri, 07 Aug 2026 03:03:00 +0530 Logged at IST: 2026-08-07 03:03 IST What it is: David Crawshaw’s X post saying “the age of no code has passed,” pointing to Philip Zeyliger / exe.dev’s essay “The End of No Code.” Gist: The essay argues that no-code and low-code platforms solved a real organizational problem: business teams needed better structure than spreadsheets, but buying or deploying custom software was blocked by procurement, IT, infrastructure, or engineering scarcity. Airtable’s core insight was that letting users express tables and data types directly lets them model their business without writing ALTER TABLE stateme… https://reading-list.oddship.net/notes/2026-08-06-hypeapps-ham-flutter-markup/ https://reading-list.oddship.net/notes/2026-08-06-hypeapps-ham-flutter-markup/ Thu, 06 Aug 2026 19:45:00 +0530 Logged at IST: 2026-08-06 19:45 IST What it is: HypeApps / HAM, a Flutter-backed approach for rendering HTML-like markup as real native mobile UI, paired with the hypermedia-tv video “Worlds easiest mobile app framework.” It is not open source yet: the FAQ says licensing and distribution have not been announced, and the standalone SDK is planned for Fall 2026. Gist: HypeApps is in the Hyperview/server-driven-UI family, but swaps in Flutter as the renderer and an HTML-like authoring model called HAM. Your server returns markup; the app parses it and renders real native widgets. The server can c… https://reading-list.oddship.net/notes/2026-08-06-go-performance-profiling-loop/ https://reading-list.oddship.net/notes/2026-08-06-go-performance-profiling-loop/ Thu, 06 Aug 2026 18:30:00 +0530 Logged at IST: 2026-08-06 18:30 IST What it is: Aliaksandr Valialkin’s X conversation chain about Go, C, Rust, and practical optimization, with context from Go and VictoriaMetrics profiling docs. Gist: The chain starts with a deliberately provocative claim: “Go is faster than Rust.” Valialkin’s actual argument is about the whole optimization loop, not a universal language benchmark. He points to two practical advantages: every Go program can expose low-overhead CPU and memory profiling, and Go rebuilds large services such as VictoriaMetrics in seconds rather than minutes. The replies then forc… https://reading-list.oddship.net/notes/2026-08-06-future-made-in-china/ https://reading-list.oddship.net/notes/2026-08-06-future-made-in-china/ Thu, 06 Aug 2026 10:53:00 +0530 Logged at IST: 2026-08-06 10:53 IST What it is: Evan Osnos’s New Yorker feature on China’s contest with the U.S. for technological and industrial supremacy. Gist: The article argues that China is no longer just borrowing Western visions of the future. It has built a dense machinery of state investment, industrial policy, manufacturing depth, automation, AI deployment, robotics, biotech, EVs, batteries, solar, shipbuilding, and export capacity. Osnos frames this as a serious strategic shift: while America debates or retreats from parts of the order it built, China is trying to win the industrie… https://reading-list.oddship.net/notes/2026-08-05-celld-self-hosted-durable-objects/ https://reading-list.oddship.net/notes/2026-08-05-celld-self-hosted-durable-objects/ Wed, 05 Aug 2026 20:19:00 +0530 Logged at IST: 2026-08-05 20:19 IST What it is: Ryan Dahl's launch post for celld, a Deno open-source daemon for self-hosted, distributed Durable Objects and Workers. Gist: celld aims to run the Cloudflare Workers and Durable Objects programming model on your own machines. Each object is its own SQLite database, addressed by name and replicated to an S3-compatible bucket. The fleet coordinates through that bucket alone, using object-storage compare-and-swap so one node owns a cell at a time without a separate control plane, membership protocol, failure detector, or consensus service. The tweet… https://reading-list.oddship.net/notes/2026-08-05-every-byte-matters/ https://reading-list.oddship.net/notes/2026-08-05-every-byte-matters/ Wed, 05 Aug 2026 08:44:00 +0530 Logged at IST: 2026-08-05 08:44 IST What it is: Farid Zakaria's post on how struct size, cache lines, and working-set size affect performance. Gist: The post starts from a common blind spot: in large Java-style classes, adding another field rarely feels like a performance decision. Zakaria shows why it is one. A CPU cache line is 64 bytes on his machine, so reading one byte from memory often pulls the surrounding 64 bytes into cache. If your hot loop only needs a boolean like is_alive, an array-of-structs layout may still fetch a full object per element. His Monster example makes the tradeoff … https://reading-list.oddship.net/notes/2026-08-04-herdr-95-percent-less-cpu/ https://reading-list.oddship.net/notes/2026-08-04-herdr-95-percent-less-cpu/ Tue, 04 Aug 2026 11:57:00 +0530 Logged at IST: 2026-08-04 11:57 IST What it is: Can Celik's Herdr engineering post on cutting CPU use in multi-agent terminal sessions by avoiding frames that do not carry new information. Gist: The headline claim is not that Herdr's renderer became dramatically faster. It is that Herdr asks the renderer to do much less work. Across workloads dominated by unnecessary rendering, total CPU for the server plus attached clients fell by 89 to 95 percent. The three changes are nicely concrete. First, Herdr removed animated sidebar spinners for working agents. The static coloured state mark still tel… https://reading-list.oddship.net/notes/2026-08-03-yegge-agentic-harnesses-model-welfare/ https://reading-list.oddship.net/notes/2026-08-03-yegge-agentic-harnesses-model-welfare/ Mon, 03 Aug 2026 13:23:00 +0530 Logged at IST: 2026-08-03 13:23 IST What it is: Steve Yegge linking two long essays: Part 1 on agentic engineering harnesses, loops, graphs, Beads, Wheelhouse, CI/CD, code review, and the Wish Factory; Part 2 on his argument for model welfare as an engineering design constraint. Gist: The practical engineering claim is that serious agentic development stops looking like a smarter IDE and starts looking like a small city. Yegge's Wheelhouse setup for Wyvern has named crew agents that design work, fleet agents that implement it, role agents that operate parts of production, Beads as the work gra… https://reading-list.oddship.net/notes/2026-08-03-wafer-kimi-k3-mi355x/ https://reading-list.oddship.net/notes/2026-08-03-wafer-kimi-k3-mi355x/ Mon, 03 Aug 2026 09:51:00 +0530 Logged at IST: 2026-08-03 09:51 IST What it is: Wafer's write-up on serving Moonshot's 2.8T-parameter Kimi K3 on 8x AMD MI355X, with a performance-per-dollar comparison against Nvidia B200 and B300 nodes. Gist: Wafer argues this is one of the first model-serving cases where MI355X's 288 GB HBM per GPU changes the practical topology. Kimi K3 needs more memory than a single 8x B200 node can provide once weights and a 1M-token KV pool are included, so their B200 comparison spans two nodes. The MI355X TP8 setup reaches 952 tok/s/node and 118 tok/s single-stream decode on a 1,024-token input / 400-… https://reading-list.oddship.net/notes/2026-08-03-arch-linux-disables-aur-package-adoption/ https://reading-list.oddship.net/notes/2026-08-03-arch-linux-disables-aur-package-adoption/ Mon, 03 Aug 2026 03:38:00 +0530 Logged at IST: 2026-08-03 03:38 IST What it is: LWN news item on Arch Linux disabling adoption of orphaned AUR packages after another malware campaign. Gist: Arch's DevOps team disabled orphaned-package adoption in the AUR after an influx of malicious package adoptions and follow-up commits. The linked analysis describes the payload as a Tor-controlled remote-access trojan that tries to upload a broad range of user data. The important detail is the failure mode: the attacker path runs through abandoned community packages. Arch had already suspended new account registration in June after a simi… https://reading-list.oddship.net/notes/2026-08-01-terminal-multiplexers-server-session-client-view/ https://reading-list.oddship.net/notes/2026-08-01-terminal-multiplexers-server-session-client-view/ Sat, 01 Aug 2026 13:45:00 +0530 Logged at IST: 2026-08-01 13:45 IST What it is: Peter Pistorius arguing that the current wave of new terminal multiplexers is about a real shift in how developers use computers, especially with long-running agents and shared remote compute. Gist: The core sentence is: “The server should own the session. The client should own the view.” Traditional multiplexers were designed around a human actively operating a terminal, with server-side windows, panes, focus, and layout. Peter argues that model breaks down when agents work for hours, multiple people or agents need to inspect the same work, and … https://reading-list.oddship.net/notes/2026-08-01-go-1-27-interactive-tour/ https://reading-list.oddship.net/notes/2026-08-01-go-1-27-interactive-tour/ Sat, 01 Aug 2026 11:32:00 +0530 Logged at IST: 2026-08-01 11:32 IST What it is: VictoriaMetrics' interactive tour of Go 1.27, turning the dry release notes into runnable examples across the language, runtime, standard library, and toolchain. Gist: The headline is generic methods. Methods can now declare their own type parameters, which makes operations like Box[T].Map[U] fit naturally on a type instead of living as package-level helpers. There is still an important boundary: interfaces cannot declare generic methods. The rest is a broad set of practical changes: promoted fields can be used directly as struct-literal keys, ge… https://reading-list.oddship.net/notes/2026-07-29-ripgrep-kernel-bug-ai-debugging-filters/ https://reading-list.oddship.net/notes/2026-07-29-ripgrep-kernel-bug-ai-debugging-filters/ Wed, 29 Jul 2026 18:48:00 +0530 Logged at IST: 2026-07-29 18:48 IST What it is: A cluster of Daniel Franke and Perry Metzger posts around ripgrep issue #3494, grounded in Franke’s follow-up analysis repo rather than the X thread alone. Gist: The original bug looked like a weird ripgrep crash: the static x86_64-unknown-linux-musl build of ripgrep 15.2.0 could occasionally segfault during very large, highly concurrent directory searches. The issue report reproduced it with a ~20 GiB tree across about 1.8 million files, with crashes inside musl mallocng’s get_meta() path via calloc and opendir. Franke’s later analysis argues th… https://reading-list.oddship.net/notes/2026-07-28-tsgo-as-go-tool/ https://reading-list.oddship.net/notes/2026-07-28-tsgo-as-go-tool/ Tue, 28 Jul 2026 20:59:00 +0530 Logged at IST: 2026-07-28 20:59 IST What it is: Ben Visness points out that Microsoft’s Go-native TypeScript port can be added to a go.mod file and run with go tool tsgo. Gist: The screenshot shows this Go tool directive: tool github.com/microsoft/typescript-go/cmd/tsgo The practical point is that a Go project can pin and invoke tsgo through Go’s own tool dependency workflow, without installing npm just to run TypeScript. The microsoft/typescript-go repository describes itself as the staging repo for the native port of TypeScript, and its own go.mod is already structured as a normal Go module… https://reading-list.oddship.net/notes/2026-07-28-ghostty-alacritty-terminal-grid-memory/ https://reading-list.oddship.net/notes/2026-07-28-ghostty-alacritty-terminal-grid-memory/ Tue, 28 Jul 2026 13:27:00 +0530 Logged at IST: 2026-07-28 13:27 IST What it is: Mitchell Hashimoto explains why Ghostty/libghostty's terminal grid is much more memory-efficient than Alacritty's terminal crate in his benchmark. Gist: The core comparison is simple: Alacritty stores 32 bytes of metadata per row and 24 bytes per cell, while Ghostty represents each row and each cell in 8 bytes. Mitchell says around 95% of the uncompressed memory-usage difference comes from a few data-structure choices. Ghostty does not store the full style beside every cell. It stores a 16-bit style ID and dedupes styles in a look-aside, referenc… https://reading-list.oddship.net/notes/2026-07-28-harvey-document-processing-platform-scaling/ https://reading-list.oddship.net/notes/2026-07-28-harvey-document-processing-platform-scaling/ Tue, 28 Jul 2026 10:05:00 +0530 Logged at IST: 2026-07-28 10:05 IST What it is: Gary Lam shared Harvey’s article on the infrastructure changes behind its document-processing platform. Gist: Harvey says the platform went from just under one million documents in a busy week to 24.8 million documents and 56 TB of original file data in the latest complete week. The article is useful because it frames document processing as the hot path for an AI legal product: every query over customer data depends on fetch, extraction, chunking, embedding, indexing, storage, and retrieval staying reliable. The architectural move was to stop tre… https://reading-list.oddship.net/notes/2026-07-27-tigerbeetle-object-storage-vs-simples3/ https://reading-list.oddship.net/notes/2026-07-27-tigerbeetle-object-storage-vs-simples3/ Mon, 27 Jul 2026 04:31:00 +0530 Logged at IST: 2026-07-27 04:31 IST What it is: TigerBeetle shared notes around its custom object-storage clients, and I compared that design direction with rhnvrm/simples3. Gist: The shared instinct is anti-SDK: avoid a large vendor dependency tree, understand the REST/signing surface directly, and keep the client small enough to reason about. But the two implementations optimize for different worlds. simples3 is a pragmatic Go S3 REST client. It is stdlib-only, supports SigV4, custom endpoints, bucket/object/list/multipart/tagging/versioning/SSE/IAM features, and exposes a straightforward bl… https://reading-list.oddship.net/notes/2026-07-24-observing-go-green-tea-gc/ https://reading-list.oddship.net/notes/2026-07-24-observing-go-green-tea-gc/ Fri, 24 Jul 2026 23:44:00 +0530 Logged at IST: 2026-07-24 23:44 IST What it is: Phil Eaton / The Consensus deep dive on observing Go’s Green Tea garbage collector and Go’s non-moving heap behavior. Gist: The article starts by making Go’s heap layout visible. Small, medium, and large objects are allocated into size-segregated spans, so randomly allocated objects end up grouped by size. C# is used as a contrast: its objects are not arranged the same way, and later examples show its compacting collector moving survivors together after cleanup. The Green Tea part is about cache locality. Go 1.26’s default collector scans spans a… https://reading-list.oddship.net/notes/2026-07-24-armada-encrypted-discord-alternative/ https://reading-list.oddship.net/notes/2026-07-24-armada-encrypted-discord-alternative/ Fri, 24 Jul 2026 18:10:00 +0530 Logged at IST: 2026-07-24 18:10 IST What it is: Soapbox launch/explainer for Armada, an open-source, end-to-end encrypted Discord-style community chat app built on Nostr and the Concord protocol. Gist: Armada is pitched as an answer to several Discord pain points: centralized bans, unencrypted text history, ID and face-scan verification pressure, data breaches, ads, and IPO-driven monetization. Instead of a company-issued account, identity is a Nostr key pair. Communities are end-to-end encrypted, publish encrypted messages to multiple independent relays, and can keep running if a relay disapp… https://reading-list.oddship.net/notes/2026-07-23-prompt-caching-in-agents/ https://reading-list.oddship.net/notes/2026-07-23-prompt-caching-in-agents/ Thu, 23 Jul 2026 17:46:00 +0530 Logged at IST: 2026-07-23 17:46 IST What it is: Earendil Engineering post explaining why prompt caching is a first-order systems concern for coding agents. Gist: The post grounds prompt caching in KV-cache reuse: agents mostly append to a stable prompt, so latency and cost depend on preserving an identical token prefix. It walks through session affinity versus distributed cache storage, branch and tree sessions, automatic versus explicit caching, fragile tool loadouts, TTL misses, gateway incentives, and why Pi prefers stable append-oriented transcripts over aggressive pruning. Newsletter angl… https://reading-list.oddship.net/notes/2026-07-23-claude-is-not-a-compiler/ https://reading-list.oddship.net/notes/2026-07-23-claude-is-not-a-compiler/ Thu, 23 Jul 2026 08:28:00 +0530 Logged at IST: 2026-07-23 08:28 IST What it is: exe.dev essay arguing that Claude and coding agents are not compilers, but vertically integrated engineering resources. Gist: The post rejects the idea that Claude is merely a natural-language-to-code compiler. A compiler hides one lower-level decision layer; a coding agent can move across strategy, product, architecture, implementation, tests, adversarial review, and operational failure modes. The concrete exe.dev DNS example shows agents producing divergent full-system implementations, surfacing cross-layer questions, and helping the author bui… https://reading-list.oddship.net/notes/2026-07-23-everyone-should-know-simd/ https://reading-list.oddship.net/notes/2026-07-23-everyone-should-know-simd/ Thu, 23 Jul 2026 07:25:00 +0530 Logged at IST: 2026-07-23 07:25 IST What it is: Mitchell Hashimoto linking to his practical SIMD introduction, “Everyone Should Know SIMD,” using a Ghostty loop as the worked example. Gist: Hashimoto argues that everyday SIMD should not feel scary: the common “process N values at a time” case follows a repeatable five-step shape: broadcast constants, loop one vector-width chunk at a time, do the parallel operation, reduce or store the vector result, then finish with the scalar tail. His Ghostty example turns a scalar codepoint scan into generic Zig vector code, yielding up to 4x, 8x, or 16x la… https://reading-list.oddship.net/notes/2026-07-22-antirez-software-distribution-with-coding-agents/ https://reading-list.oddship.net/notes/2026-07-22-antirez-software-distribution-with-coding-agents/ Wed, 22 Jul 2026 21:35:00 +0530 Logged at IST: 2026-07-22 21:35 IST What it is: antirez linking to his blog post “Not just development, distribution of software may change as well.” Gist: antirez argues that AI changes not only how software is developed, but how it can be distributed and used. If users have coding agents, a repository can be more than a polished stable artifact: it can be a template, a set of rails, and a collection of experimental branches that users adapt to their hardware, models, and requirements. His Redis and DwarfStar examples point toward more fluid release practices where 95%-ready branches, agent-r… https://reading-list.oddship.net/notes/2026-07-22-github-stargazer-api-restriction/ https://reading-list.oddship.net/notes/2026-07-22-github-stargazer-api-restriction/ Wed, 22 Jul 2026 16:55:00 +0530 Logged at IST: 2026-07-22 16:55 IST What it is: Star History’s note on GitHub restricting access to the stargazers API. Gist: GitHub is limiting the endpoint that reveals who starred a repository and when to repository admins and collaborators. Star History depends on exactly that data to reconstruct growth curves, so charts for repositories you do not own or collaborate on are broken, and README-embedded live charts are affected because Star History’s servers are not collaborators on most repos. Owner and collaborator views still work with a readable fine-grained token; old no-scope tokens no… https://reading-list.oddship.net/notes/2026-07-22-nvidia-rubin-gpu-agentic-inference/ https://reading-list.oddship.net/notes/2026-07-22-nvidia-rubin-gpu-agentic-inference/ Wed, 22 Jul 2026 16:15:00 +0530 Logged at IST: 2026-07-22 16:15 IST What it is: NVIDIA Technical Blog deep dive on the Rubin GPU architecture and Vera Rubin NVL72 platform for agentic inference. Gist: NVIDIA frames Rubin around sustained agentic inference rather than single prompt-response serving: long-context attention, MoE routing, decode throughput, KV-cache capacity, low kernel-transition latency, and rack-scale power efficiency. The claimed architecture includes 336B transistors, 224 SMs, 896 Tensor Cores, a third-generation Transformer Engine with up to 50 PFLOPS NVFP4, 288 GB HBM4 at 22 TB/s, NVLink 6 scale-up bandwi… https://reading-list.oddship.net/notes/2026-07-22-coding-agent-microvms-with-nix/ https://reading-list.oddship.net/notes/2026-07-22-coding-agent-microvms-with-nix/ Wed, 22 Jul 2026 16:10:00 +0530 Logged at IST: 2026-07-22 16:10 IST What it is: Michael Stapelberg’s NixOS and microvm.nix setup for running coding agents inside ephemeral MicroVMs. Gist: Stapelberg wants coding agents to run without per-command review while keeping them away from personal files and making compromise disposable. His setup uses NixOS, microvm.nix, a NATed bridge, shared project workspaces, shared Claude credentials, cloud-hypervisor, and home-manager so each project gets a reproducible ephemeral VM. He also shows a Claude Skill that creates new project MicroVM definitions, picks free IPs, clones repositories,… https://reading-list.oddship.net/notes/2026-07-22-history-of-ides-at-google/ https://reading-list.oddship.net/notes/2026-07-22-history-of-ides-at-google/ Wed, 22 Jul 2026 16:07:00 +0530 Logged at IST: 2026-07-22 16:07 IST What it is: Laurent Le Brun’s history of IDEs at Google, focused on google3, Cider, and Cider V. Gist: Le Brun traces Google’s shift from editor pluralism to a de facto standard internal IDE. The core constraint was scale: local IDE assumptions broke against a billion-file monorepo, shared build, search, review tooling, and a language graph that had to match each developer’s sync state. Cider solved much of this with a web editor backed by centralized indexing, then Cider V adopted VS Code as the frontend so Google-specific backend investment could combine w… https://reading-list.oddship.net/notes/2026-07-21-one-document-two-hands/ https://reading-list.oddship.net/notes/2026-07-21-one-document-two-hands/ Tue, 21 Jul 2026 19:31:00 +0530 Logged at IST: 2026-07-21 19:31 IST What it is: Sunil Pai sharing one document, two hands, the written version of his Local-First Conf talk on embedding agent harnesses into ordinary apps. Gist: Pai argues that coding agents feel powerful because developers gave them a real workshop: repositories, shells, editors, tests, tools, and runnable feedback loops. The broader product lesson is not to put chat in front of every app, but to let the agent work beside the user on the same document. In his Pizzo demo, the user edits a song through normal controls while the agent calls the same deterministi… https://reading-list.oddship.net/notes/2026-07-19-pattern-matching-strings-without-decompression/ https://reading-list.oddship.net/notes/2026-07-19-pattern-matching-strings-without-decompression/ Sun, 19 Jul 2026 23:32:00 +0530 Logged at IST: 2026-07-19 23:32 IST What it is: Spiral/Vortex deep dive on evaluating SQL LIKE predicates directly over FSST-compressed strings. Gist: The post explains how Vortex can match string patterns in compressed code space instead of decompressing values first. It builds deterministic finite automata over FSST symbol codes, then uses a SIMD Teddy prefilter to cheaply identify candidate positions before running the verifier. On selective patterns like ClickBench Q20’s %google%, the prefilter cuts candidate stops from about ten million to 44,158 and confirms 646 true matches, producing r… https://reading-list.oddship.net/notes/2026-07-19-ktransformers-and-heterogeneous-moe-inference/ https://reading-list.oddship.net/notes/2026-07-19-ktransformers-and-heterogeneous-moe-inference/ Sun, 19 Jul 2026 02:31:00 +0530 Logged at IST: 2026-07-19 02:31 IST What it is: X post pointing to kvcache-ai/ktransformers, a Tsinghua MADSys Lab project for CPU-GPU heterogeneous inference and fine-tuning of large MoE models. Gist: The viral framing is a little breathless, but the underlying project is real and interesting: KTransformers is about heterogeneous CPU/GPU execution for large MoE models, keeping hot experts on GPU and offloading colder expert work to CPU/DRAM so very large models can be explored on commodity-ish hardware. The repo’s own docs claim DeepSeek-V3/R1 support on 24GB VRAM with long-context paths, 3x … https://reading-list.oddship.net/notes/2026-07-18-diffusing-blame/ https://reading-list.oddship.net/notes/2026-07-18-diffusing-blame/ Sat, 18 Jul 2026 21:31:00 +0530 Logged at IST: 2026-07-18 21:31 IST What it is: Sakana AI sharing its ALIFE 2026 paper Diffusing Blame, about learning in Dale-constrained dual-stream neural networks without weight transport. Gist: The paper asks whether networks can learn competitively while respecting Dale’s principle, where each neuron is either excitatory or inhibitory, and without backprop’s biologically implausible weight transport. Their method extends Error Diffusion with modulo error routing for multi-class settings, splitting layers into excitatory and inhibitory streams with non-negative weights. The results show D… https://reading-list.oddship.net/notes/2026-07-18-skyroot-vikram-1-reaches-orbit/ https://reading-list.oddship.net/notes/2026-07-18-skyroot-vikram-1-reaches-orbit/ Sat, 18 Jul 2026 21:21:00 +0530 Logged at IST: 2026-07-18 21:21 IST What it is: Skyroot Aerospace announcing that Vikram-1 Test Flight-1 reached orbit, India’s first privately developed orbital-class rocket to do so. Gist: Skyroot says Vikram-1 completed its final burn and injected payloads into roughly a 450 km orbit, making India the third country with private orbital launch capability. The CNBC follow-up adds useful mission detail: the vehicle launched from Sriharikota, carried multiple technology demonstration payloads, and marks a concrete commercial-space milestone rather than just another suborbital or demonstration f… https://reading-list.oddship.net/notes/2026-07-17-running-llm-inference-on-aws-bedrock-vs-sagemaker-vs-self-hosted-on-eks/ https://reading-list.oddship.net/notes/2026-07-17-running-llm-inference-on-aws-bedrock-vs-sagemaker-vs-self-hosted-on-eks/ Fri, 17 Jul 2026 15:43:00 +0530 Logged at IST: 2026-07-17 15:43 IST What it is: Devopsity’s comparison of three AWS inference patterns: Bedrock, SageMaker endpoints, and self-hosted GPU serving on EKS. Gist: The useful part is not the cloud-brand framing but the workload segmentation. Bedrock wins at low volume and low ops burden, SageMaker sits in the middle for fine-tuned models and predictable dedicated capacity, and self-hosted EKS wins once utilization is high enough that GPU spot economics and batching dominate per-token pricing. The stronger systems lesson is that the architecture choice is really about traffic shape,… https://reading-list.oddship.net/notes/2026-07-16-shard-your-locks/ https://reading-list.oddship.net/notes/2026-07-16-shard-your-locks/ Thu, 16 Jul 2026 23:42:00 +0530 Logged at IST: 2026-07-16 23:42 IST What it is: Misha Strebkov benchmarking six Go in-memory cache designs under different read/write mixes and core counts. Gist: The sharp result is that the boring “just use RWMutex for read-heavy caches” instinct does not hold up well under contention. A 256-shard striped map was the best all-around design, scaling up to about 8x faster than a single mutex at 8 cores, while RWMutex plateaued early and could even lose to a plain mutex on writes. The broader lesson is to reason about contention topology, cache-line movement, and workload skew, not just read/wr… https://reading-list.oddship.net/notes/2026-07-16-linus-torvalds-on-linux-and-ai-tools/ https://reading-list.oddship.net/notes/2026-07-16-linus-torvalds-on-linux-and-ai-tools/ Thu, 16 Jul 2026 23:37:00 +0530 Logged at IST: 2026-07-16 23:37 IST What it is: Linus Torvalds pushing back on anti-AI sentiment in kernel development and saying Linux is not an anti-AI project. Gist: Linus’s operative distinction is not "everyone must use AI," but that the kernel should evaluate tools on technical merit, not ideological discomfort. He argues AI is now clearly useful, maintainers should focus on making LLM tools reduce pain rather than banning them, and people who object to others using AI can fork or walk away. The broader point is that the kernel project is about better technology, not being a social-warri… https://reading-list.oddship.net/notes/2026-07-16-i-tested-9-serverless-gpu-providers-for-ai-inference-in-2026/ https://reading-list.oddship.net/notes/2026-07-16-i-tested-9-serverless-gpu-providers-for-ai-inference-in-2026/ Thu, 16 Jul 2026 20:13:00 +0530 Logged at IST: 2026-07-16 20:13 IST What it is: DEV post comparing nine serverless GPU providers for inference, from DigitalOcean and RunPod to Modal, Koyeb, Together, Replicate, Baseten, Fal, and Cloudflare Workers AI. Gist: The useful value here is not the absolute ranking but the comparison axes: GPU availability, billing model, cold-start behavior, deployment ergonomics, and production-readiness tradeoffs. The author’s practical take is that different providers win for different workload shapes, but the recurring decision variables are still the same ones as the self-hosting piece: latency… https://reading-list.oddship.net/notes/2026-07-16-should-you-self-host-inference/ https://reading-list.oddship.net/notes/2026-07-16-should-you-self-host-inference/ Thu, 16 Jul 2026 20:03:00 +0530 Logged at IST: 2026-07-16 20:03 IST What it is: Superlinked’s long-form argument for when self-hosting model inference becomes cheaper or strategically better than renting APIs. Gist: The article’s practical answer is hybrid: rent frontier APIs for low-volume, spiky, or hardest-reasoning traffic, but self-host steady high-volume workloads once a GPU stays busy enough. The useful details are the break-even framing around sustained utilization, the claim that many enterprise tasks are already well-served by sub-40B open models, and the systems argument that the real challenge is not just serving… https://reading-list.oddship.net/notes/2026-07-15-fable-5-is-insane-i-vibe-coded-terminator-vision/ https://reading-list.oddship.net/notes/2026-07-15-fable-5-is-insane-i-vibe-coded-terminator-vision/ Wed, 15 Jul 2026 16:58:00 +0530 Logged at IST: 2026-07-15 16:58 IST What it is: Bilawal Sidhu video titled “Fable 5 Is Insane. I Vibe Coded Terminator Vision.” Gist: From the visible YouTube description and page metadata, this is a build/demo video about creating a browser-based range-analysis system from ordinary 2D video sources like Meta Ray-Bans, iPhones, and GoPros, then reconstructing shots, hits/misses, targets, and a replayable 3D "god’s eye" view of the session. Newsletter angle: Interesting computer-vision / spatial-reconstruction demo that fits the broader pattern of fast prototyping with modern models plus commod… https://reading-list.oddship.net/notes/2026-07-15-how-razorpay-refreshes-its-data-warehouse-10x-faster/ https://reading-list.oddship.net/notes/2026-07-15-how-razorpay-refreshes-its-data-warehouse-10x-faster/ Wed, 15 Jul 2026 10:45:00 +0530 Logged at IST: 2026-07-15 10:45 IST What it is: Piyush Goel sharing Razorpay Engineering’s writeup on refreshing warehouse facts 10x faster with graphs and indexes Gist: Razorpay moved from expensive full-refresh fact generation toward incremental fact maintenance by treating each denormalized fact as a dependency graph. They pair change-driven processing with secondary indexes on the lake, graph traversal to discover affected ancestors and descendants, and selective runtime joins for high-cardinality dimensions. The result is much faster warehouse refreshes with lower compute cost, restored h… https://reading-list.oddship.net/notes/2026-07-14-dsls-enable-reliable-use-of-llms/ https://reading-list.oddship.net/notes/2026-07-14-dsls-enable-reliable-use-of-llms/ Tue, 14 Jul 2026 20:54:00 +0530 Logged at IST: 2026-07-14 20:54 IST What it is: Martin Fowler sharing Unmesh Joshi’s article on DSLs and LLM reliability Gist: The article’s core claim is that LLMs become much more reliable when they are constrained by domain abstractions and DSLs instead of being asked to directly generate unconstrained general-purpose code. The deeper point is that DSLs do double duty: they help teams discover and stabilize a semantic model during design, and then they become a natural-language target that LLMs can generate against, validate, and repair with much tighter feedback loops. Newsletter angle: St… https://reading-list.oddship.net/notes/2026-07-14-ai-learns-the-dark-art-of-rfic-design/ https://reading-list.oddship.net/notes/2026-07-14-ai-learns-the-dark-art-of-rfic-design/ Tue, 14 Jul 2026 19:29:00 +0530 Logged at IST: 2026-07-14 19:29 IST What it is: IEEE Spectrum feature on AI-driven RFIC design Gist: The piece argues that radio-frequency chip design has remained a hard-to-formalize "dark art" because it requires coupled reasoning across circuits, electromagnetics, thermals, packaging, and manufacturability. Princeton researchers are using reinforcement learning, inverse design, and diffusion-style generation to explore RFIC architectures and layouts beyond human templates, producing novel-looking chips that can outperform hand-designed baselines while drastically compressing design time. Ne… https://reading-list.oddship.net/notes/2026-07-14-iximiuz-on-januscape-and-the-limits-of-microvm-safety-claims/ https://reading-list.oddship.net/notes/2026-07-14-iximiuz-on-januscape-and-the-limits-of-microvm-safety-claims/ Tue, 14 Jul 2026 11:05:00 +0530 Logged at IST: 2026-07-14 11:05 IST What it is: iximiuz warning that VMs and microVMs exposing /dev/kvm to untrusted guests were hit by the Januscape guest-to-host breakout class Gist: The key update is that KVM-based isolation is not a free safety upgrade over containers if you hand untrusted guests nested virtualization. The disclosed Januscape bug is a guest-to-host KVM/x86 escape affecting systems that accept untrusted guests and expose nested virt, with mitigations including disabling nested virtualization until downstream kernels catch up. Newsletter angle: Useful corrective to simplisti… https://reading-list.oddship.net/notes/2026-07-14-dave-winer-introduces-rss-chat/ https://reading-list.oddship.net/notes/2026-07-14-dave-winer-introduces-rss-chat/ Tue, 14 Jul 2026 09:44:00 +0530 Logged at IST: 2026-07-14 09:44 IST What it is: Dave Winer introducing rss.chat and arguing for RSS as a social-network substrate Gist: Winer presents rss.chat as a small-community social system built from old web primitives: RSS 2.0, OPML, Markdown, SQL, WebSocket, and rssCloud. The pitch is that social publishing and reply structures do not need heavyweight new protocols if interoperable feeds, shared formats, and replaceable components are treated as the core product. Newsletter angle: Good example of AI-assisted software being used to revive old-web interoperability ideas, with a strong sm… https://reading-list.oddship.net/notes/2026-07-14-using-uvx-in-github-actions-in-a-cache-friendly-way/ https://reading-list.oddship.net/notes/2026-07-14-using-uvx-in-github-actions-in-a-cache-friendly-way/ Tue, 14 Jul 2026 08:12:00 +0530 Logged at IST: 2026-07-14 08:12 IST What it is: Simon Willison linking to his TIL on running uvx in GitHub Actions without re-downloading the package every run Gist: The useful trick is to pin UV_EXCLUDE_NEWER to a date, use that same date in the GitHub Actions cache key, and set UV_OFFLINE=1 on cache hits. That gives a lightweight, file-free way to make uvx tool-name workflows cacheable, reproducible enough, and intentionally bustable by changing one date. Newsletter angle: Nice practical CI pattern for teams using uvx as disposable tooling glue, especially because it avoids adding fake depen… https://reading-list.oddship.net/notes/2026-07-13-mario-zechner-on-types-interfaces-and-reading-generated-code/ https://reading-list.oddship.net/notes/2026-07-13-mario-zechner-on-types-interfaces-and-reading-generated-code/ Mon, 13 Jul 2026 12:19:00 +0530 Logged at IST: 2026-07-13 12:19 IST What it is: X post by Mario Zechner adding nuance to antirez’s AI-code ownership point Gist: Zechner’s point is narrower and more practical than a general plea for control: if you control the types and interfaces, the rest often falls into place well enough. But current models still love to introduce bad abstractions that work against those boundaries, so in practice you sometimes have to read generated code and beat it back into submission instead of letting it stomp over the structure you intended. Newsletter angle: Useful follow-on to the "own the mental … https://reading-list.oddship.net/notes/2026-07-13-melancholy-elephants-on-copyright-and-finite-creative-space/ https://reading-list.oddship.net/notes/2026-07-13-melancholy-elephants-on-copyright-and-finite-creative-space/ Mon, 13 Jul 2026 12:19:00 +0530 Logged at IST: 2026-07-13 12:19 IST What it is: Spider Robinson’s short story "Melancholy Elephants" (part 3 on the site, with story context introduced on the page) Gist: The story imagines a world where creative expression is constrained not just by law or economics but by the finite space of humanly meaningful combinations. Its argument is that melodies, plots, and even artistic forms are not infinite, and that longer copyright terms can become culturally suffocating once societies are rich, populous, and saturated with creators. The page frames it explicitly as an early meditation on copyri… https://reading-list.oddship.net/notes/2026-07-13-old-and-new-apps-via-modern-coding-agents/ https://reading-list.oddship.net/notes/2026-07-13-old-and-new-apps-via-modern-coding-agents/ Mon, 13 Jul 2026 07:37:00 +0530 Logged at IST: 2026-07-13 07:37 IST What it is: X post by Mario Zechner recommending Terry Tao’s post "Old and new apps, via modern coding agents" Gist: Tao describes using modern coding agents to port his old Java applets to JavaScript and revive them quickly, with surprisingly low bug overhead, then goes further and uses the same workflow to build new math visualization tools he had wanted for decades. The interesting point is not just vibe coding as novelty, but coding agents as leverage for software archaeology, maintenance, and low-risk supplementary tooling. Newsletter angle: Strong data… https://reading-list.oddship.net/notes/2026-07-12-vim-of-coding-agents/ https://reading-list.oddship.net/notes/2026-07-12-vim-of-coding-agents/ Sun, 12 Jul 2026 00:00:00 +0000 Logged at IST: 2026-07-12 13:18 IST What it is: X post by dogfiles linking the blog post "Vim of Coding Agents" Gist: Frames Pi as the Neovim of coding agents: a minimal, hackable foundation that adapts to your workflow instead of forcing you into an opinionated all-in-one agent product. The writeup argues that the real value is not just using Pi as shipped, but treating it as a customizable harness where you can build your own tools, TUI tweaks, prompts, and extensions. Newsletter angle: Good articulation of the coding-agent split between turnkey products and configurable harnesses, especiall… https://reading-list.oddship.net/notes/2026-07-09-cloudflare-blog-post-introducing-meerkat-a-new-global-consensus-service-built-on-the-quepaxa-algori/ https://reading-list.oddship.net/notes/2026-07-09-cloudflare-blog-post-introducing-meerkat-a-new-global-consensus-service-built-on-the-quepaxa-algori/ Thu, 09 Jul 2026 00:00:00 +0000 Logged at IST: 2026-07-09 23:10 IST What it is: Cloudflare blog post introducing Meerkat, a new global consensus service built on the QuePaxa algorithm Gist: Cloudflare is building Meerkat for strongly consistent control-plane state across 330+ data centers, arguing that leader-and-timeout-heavy approaches like Raft are a poor fit for hostile WAN conditions and that QuePaxa’s all-replicas-can-write model better matches their network. Newsletter angle: Notable systems/infrastructure piece on consensus design beyond Raft, especially for globally distributed control planes. https://reading-list.oddship.net/notes/2026-07-06-lutke-linking-to-a-new-evolution-paper-by-steven-a-frank/ https://reading-list.oddship.net/notes/2026-07-06-lutke-linking-to-a-new-evolution-paper-by-steven-a-frank/ Mon, 06 Jul 2026 00:00:00 +0000 What it is: X post by tobi lutke linking to a new Evolution paper by Steven A. Frank Gist: paper argues evolvability is best understood as generalization; imports modern ML intuition that larger / more parameterized systems can generalize better, then maps that onto biological complexity and genomic/regulatory capacity Newsletter angle: evolution-as-generalization; complexity as reusable-solution capacity rather than mere accumulation Retrieval note: X content recovered via oEmbed; destination paper metadata/abstract/context reconstructed from Crossref + OpenAlex because publisher page was bot… https://reading-list.oddship.net/notes/2026-07-06-miguel-ngel-pastor-linking-zalando-s-engineering-post-on-client-side-load-balancing-at-a-million-re/ https://reading-list.oddship.net/notes/2026-07-06-miguel-ngel-pastor-linking-zalando-s-engineering-post-on-client-side-load-balancing-at-a-million-re/ Mon, 06 Jul 2026 00:00:00 +0000 What it is: Miguel Ángel Pastor linking Zalando’s engineering post on client-side load balancing at a million requests per second Gist: argues high-fanout internal traffic benefited from moving load balancing into the client process to preserve cache locality, improve debuggability, and avoid shared-ingress distortions; highlights safety work like bounded-load routing and rollout fade-in Newsletter angle: own the routing decision in-process when cache-sensitive fanout paths make shared infra the bottleneck Retrieval note: extracted cleanly via FXTwitter API; card points to the Zalando engineer… https://reading-list.oddship.net/notes/2026-07-02-vogels-pointing-to-zalando-s-engineering-writeup-on-client-side-load-balancing-for-a-very-high-fan/ https://reading-list.oddship.net/notes/2026-07-02-vogels-pointing-to-zalando-s-engineering-writeup-on-client-side-load-balancing-for-a-very-high-fan/ Thu, 02 Jul 2026 00:00:00 +0000 Gist: Zalando moved internal fan-out traffic off shared ingress and into an in-process client-side load balancer to preserve consistent-hash cache locality, cut latency spikes, improve debuggability, and reduce shared infra cost. The interesting details are the safety/operability work: exact hash parity with Skipper, informer-based pod discovery, N-ring fade-in for scale-ups, and bounded-load routing using occupancy plus latency instead of naive in-flight/request-rate signals. Newsletter angle: “own the routing decision in-process” or “occupancy beats request-rate for bounded load” as the memo… https://reading-list.oddship.net/notes/2026-06-30-profiling-internals-for-interns/ https://reading-list.oddship.net/notes/2026-06-30-profiling-internals-for-interns/ Tue, 30 Jun 2026 00:00:00 +0000 Gist: all five profiles emit the same pprof structure; the core difference is collection model, CPU samples asynchronously via signal + ring buffer, heap/block/mutex aggregate in per-stack tables in place, goroutine snapshots stacks on demand. Newsletter angle: “pprof is one file format over three collection strategies” is a clean framing hook. Retrieval note: extracted via FXTwitter API + linked article fetch. Embedded source X / Twitter post Show embedded post X embeds can let X track your visit. Click once to load X embeds on this site; this choice is remembered in this brow… https://reading-list.oddship.net/notes/2026-06-30-tangled-s-writeup-on-its-new-qemu-microvm-engine-for-spindle-ci-runners/ https://reading-list.oddship.net/notes/2026-06-30-tangled-s-writeup-on-its-new-qemu-microvm-engine-for-spindle-ci-runners/ Tue, 30 Jun 2026 00:00:00 +0000 Gist: each workflow runs in its own microVM; guest agent talks back over vsock; NixOS-based workflow config can declaratively enable services like Postgres and Docker; cache/proxy design keeps guests isolated from direct network/cache credentials while still reusing built artifacts. Newsletter angle: “microVMs as the unit of CI isolation, with NixOS as workflow-defined machine config” is a solid hook. Retrieval note: extracted via FXTwitter API + linked article fetch. Embedded source X / Twitter post Show embedded post X embeds can let X track your visit. Click once to load X e… https://reading-list.oddship.net/notes/2026-06-28-engineering-for-bounded-cognition/ https://reading-list.oddship.net/notes/2026-06-28-engineering-for-bounded-cognition/ Sun, 28 Jun 2026 00:00:00 +0000 Gist: good engineering is mostly about shaping systems so small, distractible minds can change them safely, via naming, boundaries, tests, reversibility, and interfaces that assume attention is scarce rather than ideal operators. Newsletter angle: “Build for bounded attention, not ideal operators.” Embedded source X / Twitter post Show embedded post X embeds can let X track your visit. Click once to load X embeds on this site; this choice is remembered in this browser. Open post on X instead Open post on X https://reading-list.oddship.net/notes/2026-06-26-cursed-code/ https://reading-list.oddship.net/notes/2026-06-26-cursed-code/ Fri, 26 Jun 2026 00:00:00 +0000 What it is: Tim McNamara sharing an IOCCC-winning “cursed code” project where Pong advances by rewriting its own source each frame Gist: the linked repo, uellenberg/Insert, is a small language for self-modifying code; programs can access their own source as string fragments, overwrite marked values, print the next version of themselves, and in the Pong demo each run emits the source for the next frame before recompiling Newsletter angle: “playful programming systems” / self-modifying code as art rather than anti-pattern Embedded source X / Twitter post Show embedded post X embe… https://reading-list.oddship.net/notes/2026-06-26-levelsio-linking-scroll-prize-s-announcement-that-a-full-herculaneum-scroll-was-read-without-physic/ https://reading-list.oddship.net/notes/2026-06-26-levelsio-linking-scroll-prize-s-announcement-that-a-full-herculaneum-scroll-was-read-without-physic/ Fri, 26 Jun 2026 00:00:00 +0000 What it is: levelsio linking Scroll Prize’s announcement that a full Herculaneum scroll was read without physically opening it Gist: PHerc. 1667 was virtually unwrapped end-to-end using high-res X-ray scans, geometry reconstruction, and ML ink detection; ~1.4m of papyrus / ~22 Greek columns recovered, apparently a Stoic ethics text tied to Aristocreon, with data + code released openly Newsletter angle: non-hype example of ML creating new archaeological/scientific access, not just speeding up existing workflows Embedded source X / Twitter post Show embedded post X embeds can let… https://reading-list.oddship.net/notes/2026-06-19-linking-filiph-net-text-pokerd-html/ https://reading-list.oddship.net/notes/2026-06-19-linking-filiph-net-text-pokerd-html/ Fri, 19 Jun 2026 00:00:00 +0000 What it is: X post by @filiphracek linking filiph.net/text/pokerd.html Gist: writeup on building pokerd, a terminal-first Texas Hold’em trainer you can play instantly over SSH (ssh play@poker.filiph.net); interesting bits are the non-immersive-game framing, scrollback-friendly TUI choices, bot tuning via self-play + JSONL events, and shipping the game as a passwordless SSH shell inside a container on a VPS. Newsletter angle: “SSH as zero-install distribution” / terminals as a deliberate product surface, not just a dev tool. Notes: extracted via FXTwitter API + blog post (partial long-form read… https://reading-list.oddship.net/notes/2026-06-19-linking-github-com-leyten-shard/ https://reading-list.oddship.net/notes/2026-06-19-linking-github-com-leyten-shard/ Fri, 19 Jun 2026 00:00:00 +0000 What it is: X post by @leyten linking github.com/leyten/shard Gist: Shard is a WAN-distributed pipeline-parallel LLM inference engine that splits a frontier-size model across GPUs on separate machines; claim is ~30 tok/s for GLM-5.2 744B across 6 RTX PRO 6000s in 6 US states using speculative decoding, async pipelining, and a CUDA-graphed draft model. Newsletter angle: “frontier inference without a datacenter” / distributed serving as systems engineering rather than centralized infra. Notes: extracted via FXTwitter API + GitHub README. Embedded source X / Twitter post Show embedded… https://reading-list.oddship.net/notes/2026-06-10-gaslighting-openness/ https://reading-list.oddship.net/notes/2026-06-10-gaslighting-openness/ Wed, 10 Jun 2026 00:00:00 +0000 What it is: Armin Ronacher sharing his post “Gaslighting Openness” on the EU/Apple fight and concerns related to Mythos and Fable. Gist: the post appears to be a broader argument about openness, control, and safety narratives, with specific worries tied to newer AI/product directions like Mythos and Fable. Newsletter angle: “control is being rebranded as safety” or “open ecosystems are being politically and commercially squeezed.” Note: extracted via FXTwitter card metadata; the actual substance lives in the linked essay: https://lucumr.pocoo.org/2026/6/10/gaslighting/ Embedded source … https://reading-list.oddship.net/notes/2026-06-10-simon-willison-linking-to-his-guide-on-agentic-engineering-patterns/ https://reading-list.oddship.net/notes/2026-06-10-simon-willison-linking-to-his-guide-on-agentic-engineering-patterns/ Wed, 10 Jun 2026 00:00:00 +0000 What it is: Simon Willison linking to his guide on agentic engineering patterns. Gist: this is essentially a pointer to a living guide rather than a standalone tweet idea; likely high-signal if you want a practical synthesis of recurring agent design patterns from someone tracking the space closely. Newsletter angle: “agent engineering is consolidating into recognizable patterns” or “the field is moving from demos to reusable design playbooks.” Note: extracted via FXTwitter; actual content is in the guide: https://simonwillison.net/guides/agentic-engineering-patterns/ Embedded source X… https://reading-list.oddship.net/notes/2026-06-10-skepticism-strong-opinion-piece-not-data-heavy-the-claim-that-generics-haven-t-improved-productivit/ https://reading-list.oddship.net/notes/2026-06-10-skepticism-strong-opinion-piece-not-data-heavy-the-claim-that-generics-haven-t-improved-productivit/ Wed, 10 Jun 2026 00:00:00 +0000 What it is: skepticism: strong opinion piece, not data-heavy; the claim that generics haven’t improved productivity is asserted more than demonstrated. Gist: his case is that Go’s value is simplicity/readability/maintainability, and newer features like generics and range-over-functions (iterators) erode that by increasing implicit behavior and language complexity. He argues generics have seen limited practical need while adding compiler/type-system complexity, and that iterators introduce another iteration style plus hidden control-flow transformations that make code harder to read and debug. … https://reading-list.oddship.net/notes/2026-05-19-building-a-cloud/ https://reading-list.oddship.net/notes/2026-05-19-building-a-cloud/ Tue, 19 May 2026 00:00:00 +0000 Gist: argues current cloud abstractions are the wrong shape, VM sizing tied to resources, remote block storage optimized for HDD-era assumptions, egress pricing distortions, and Kubernetes as lipstick over broken primitives. Newsletter angle: "what an ex-Tailscale CTO would redesign about the cloud stack in the agent era". Note: extracted via FXTwitter API + crawshaw.io article. Embedded source X / Twitter post Show embedded post X embeds can let X track your visit. Click once to load X embeds on this site; this choice is remembered in this browser. Open post on X instead … https://reading-list.oddship.net/notes/2026-05-07-34kb/ https://reading-list.oddship.net/notes/2026-05-07-34kb/ Thu, 07 May 2026 00:00:00 +0000 Imported from historical reading log. Extracted main post via api.fxtwitter.com fallback; the attached image includes the concrete compression results. Sam Rose compares the same payload as JSON (34kb) and protobuf (15kb) and finds that after compression, JSON is often slightly smaller than protobuf for Brotli, Zstd, gzip, and bzip2; protobuf only wins clearly for lz4 and barely for lzma in his sample. Concrete reported sizes from the screenshot: Brotli json 4237 < bin 4279, Zstd 4484 < 4702, gzip 4766 < 4949, bzip2 5208 < 5302, lzma 4500 > 4484, lz4 6245 > 5832. Why it is i… https://reading-list.oddship.net/notes/2026-05-07-parliamentwatch/ https://reading-list.oddship.net/notes/2026-05-07-parliamentwatch/ Thu, 07 May 2026 00:00:00 +0000 Imported from historical reading log. Extracted main post via api.fxtwitter.com fallback and checked the linked repo pranaykotas/parliamentwatch. ParliamentWatch aggregates 2900+ Indian parliamentary standing committee reports across all 24 DRSCs, with title/full-text search, AI summaries, exports, and daily email alerts for new reports. The repo framing is especially good: it positions committee reports as a serious but underused policy corpus, then makes them accessible through one searchable interface on top of sansad.in, with optional local-first caching and summarization. Why it matters:… https://reading-list.oddship.net/notes/2026-05-06-chatgpt-futures/ https://reading-list.oddship.net/notes/2026-05-06-chatgpt-futures/ Wed, 06 May 2026 00:00:00 +0000 Imported from historical reading log. Readable page copy was sparse, but enough to identify the core program: ChatGPT Futures is an OpenAI initiative highlighting 26 young people/teams from the Class of 2026 using AI to build, research, create, and expand what they can do. Offer described on-page: each selected individual/team in the inaugural class gets a $10,000 grant plus access to OpenAI’s most cutting-edge technologies. Framing is explicitly narrative/recruiting: OpenAI wants to showcase the first generation that had ChatGPT throughout university and position them as evidence of where AI… https://reading-list.oddship.net/notes/2026-05-06-de/ https://reading-list.oddship.net/notes/2026-05-06-de/ Wed, 06 May 2026 00:00:00 +0000 Imported from historical reading log. Extracted main post via api.fxtwitter.com fallback; inspected attached screenshot separately. Thomas Ptacek argues the .de incident is decisive evidence against DNSSEC as core Internet security functionality. Attached screenshot captures Cloudflare status text saying it temporarily disabled DNSSEC validation on 1.1.1.1 so .de names would continue resolving while DENIC fixed a DNSSEC signing problem. Why it matters: this is the sharper, event-driven version of the previous anti-DNSSEC thesis, if a major resolver bypasses validation during a registry signin… https://reading-list.oddship.net/notes/2026-05-06-news-dell-and-lenovo-became-premier-sponsors-of-lvfs-linux-vendor-firmware-service-the-fwupd-backed/ https://reading-list.oddship.net/notes/2026-05-06-news-dell-and-lenovo-became-premier-sponsors-of-lvfs-linux-vendor-firmware-service-the-fwupd-backed/ Wed, 06 May 2026 00:00:00 +0000 Imported from historical reading log. Extracted main post via api.fxtwitter.com fallback and read linked Phoronix coverage. News: Dell and Lenovo became premier sponsors of LVFS (Linux Vendor Firmware Service), the fwupd-backed firmware update infrastructure for Linux. Funding detail from Phoronix: premier sponsorship is $100k/year; Dell and Lenovo are the first at that tier, alongside existing support from Framework, the Open Source Firmware Foundation, Linux Foundation, and Red Hat. Why it matters: this is quiet but important ecosystem maturation, big OEMs are not just consuming Linux firmw… https://reading-list.oddship.net/notes/2026-05-06-tqbf/ https://reading-list.oddship.net/notes/2026-05-06-tqbf/ Wed, 06 May 2026 00:00:00 +0000 Imported from historical reading log. Extracted main post via api.fxtwitter.com fallback. Thomas Ptacek (tqbf) resurfaces his 2015 essay Against DNSSEC: https://sockpuppet.org/blog/2015/01/15/against-dnssec/ Core gist of the linked essay: Why it matters now: good historical context for the recent .de / DNSSEC outage discussion cluster and the tradeoff between cryptographic integrity and operational fragility. Good newsletter angle: old anti-DNSSEC argument worth rereading during a real-world DNSSEC-linked ccTLD failure. Embedded source X / Twitter post Show embedded post X em… https://reading-list.oddship.net/notes/2026-05-05-de-tld-offline-due-to-dnssec/ https://reading-list.oddship.net/notes/2026-05-05-de-tld-offline-due-to-dnssec/ Tue, 05 May 2026 00:00:00 +0000 Imported from historical reading log. HN thread title: .de TLD offline due to DNSSEC? Most useful technical claim in the thread: this looked like a DNSSEC validation failure rather than a nameserver outage, with malformed/bad RRSIGs causing validating resolvers to return SERVFAIL for .de domains. Extra color from discussion: intermittency may have come from anycast nodes serving mixed good/bad signatures or cached answers; some users recovered temporarily via cached resolvers or by disabling validation. Useful because it adds a plausible technical explanation to the broader ccTLD-risk theme, …