https://reading-list.oddship.net A curated linklog of essays, posts, papers, and notes. Tue, 25 Aug 2026 20:19:00 +0530 https://reading-list.oddship.net/notes/2026-08-25-goodharts-law-human-reward-hacking/ https://reading-list.oddship.net/notes/2026-08-25-goodharts-law-human-reward-hacking/ Tue, 25 Aug 2026 20:19:00 +0530 Logged at IST: 2026-08-25 20:19 IST What it is: Manav Rathi connecting Goodhart's law to culture/personality guardrails and human reward hacking. Gist: The linked note is only a few lines, but the point is useful: “When a measure becomes a target, it ceases to be a good measure.” Rathi's gloss is that humans reward-hack and models reward-hack for the same structural reason: optimization finds the gap between a proxy and the thing it is supposed to measure. The X post applies that to a quoted report about Anthropic asking candidates how they would feel if stock went to zero after a significant … 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-find-problems-staff-engineer/ https://reading-list.oddship.net/notes/2026-08-23-find-problems-staff-engineer/ Sun, 23 Aug 2026 18:21:00 +0530 Logged at IST: 2026-08-23 18:21 IST What it is: Lalit Maganti on how he finds problems worth solving as a Staff Engineer, especially in infrastructure and developer-tools contexts. Gist: The useful move is to treat problem discovery as ambient listening, not a scheduled “think strategically” exercise. Maganti watches the normal flow of meetings, chats, emails, and complaints, then asks follow-up questions to understand the real workflow pain behind requested solutions. He is careful not to jump on the first loud request. Problems need to accumulate: the same pain showing up across teams is str… https://reading-list.oddship.net/notes/2026-08-21-ai-consciousness-liability-trap/ https://reading-list.oddship.net/notes/2026-08-21-ai-consciousness-liability-trap/ Fri, 21 Aug 2026 01:18:00 +0530 Logged at IST: 2026-08-21 01:18 IST What it is: MIT Technology Review op-ed, “Debates over AI consciousness are a trap”. Gist: The piece argues that both “superhuman runaway AI” rhetoric and AI-rights/personhood arguments can point in the same dangerous direction: making AI systems look so advanced, autonomous, or morally separate that the companies building them can disclaim responsibility for harms. The author’s central move is to bring the debate back from philosophy to product liability. AI systems are not natural beings that independently entered society. They are corporate-built software… https://reading-list.oddship.net/notes/2026-08-19-bridgewater-ai-analyst-pat/ https://reading-list.oddship.net/notes/2026-08-19-bridgewater-ai-analyst-pat/ Wed, 19 Aug 2026 18:46:00 +0530 Logged at IST: 2026-08-19 18:46 IST What it is: A LangChain video titled How Bridgewater Built an AI Analyst That Does Hours of Expert Research in Minutes, about Bridgewater's AIA Pocket Analyst Tool, or PAT. Gist: LangChain describes PAT as an internal AI analyst deployed to hundreds of Bridgewater investors. The point is not just a chatbot over documents: it is a hedge-fund-specific analyst built around Bridgewater's proprietary data, investment methodologies, and expert investor feedback, with guardrails around how that internal knowledge is used. PAT is one component of Bridgewater's broad… https://reading-list.oddship.net/notes/2026-08-19-mathematics-in-the-age-of-ai/ https://reading-list.oddship.net/notes/2026-08-19-mathematics-in-the-age-of-ai/ Wed, 19 Aug 2026 14:33:00 +0530 Logged at IST: 2026-08-19 14:33 IST What it is: Terence Tao's essay, based on a public lecture at ICM 2026, on how mathematics should respond if AI tools become capable of research-level mathematical work. Gist: Tao deliberately does not make the paper about whether that capability will arrive. He conditions on a reasonably strong version of the AI-capability hypothesis and asks the orthogonal question: what are the actual goals, objectives, and values of mathematical research, including the implicit ones the community optimizes for in practice? The key move is to treat problem-solving as a ca… https://reading-list.oddship.net/notes/2026-08-19-ai-removing-middle-class-software-engineering/ https://reading-list.oddship.net/notes/2026-08-19-ai-removing-middle-class-software-engineering/ Wed, 19 Aug 2026 11:10:00 +0530 Logged at IST: 2026-08-19 11:10 IST What it is: Florian Herrengt's essay arguing that AI is removing the middle class of software engineering. Gist: The core claim is not "AI is bad". Herrengt says he uses AI heavily. The warning is that AI removes the speed limit from weak engineering culture. Before AI, bad decisions compounded slowly because implementation was expensive. Now someone can generate a 25,000-line PR, add services, change schemas, and ship plausible-looking functionality before anyone has built a real mental model of what changed. The output may work at first, but the debt is hi… https://reading-list.oddship.net/notes/2026-08-16-theres-no-speed-limit/ https://reading-list.oddship.net/notes/2026-08-16-theres-no-speed-limit/ Sun, 16 Aug 2026 13:11:00 +0530 Logged at IST: 2026-08-16 13:11 IST What it is: DHH shared Derek Sivers' short essay about Kimo Williams, the teacher who reset Sivers' sense of how fast he could learn before starting at Berklee. Gist: Sivers describes calling a local recording studio before college and meeting Kimo Williams, who offered to teach him two years of theory and arranging in a few lessons. The pace was intense. Kimo kept him slightly over his head, quizzed him immediately, and expected him to prove he understood each idea. That compressed instruction worked. Sivers tested out of six semesters of requirements, then… https://reading-list.oddship.net/notes/2026-08-13-keep-thinking-in-agent-work/ https://reading-list.oddship.net/notes/2026-08-13-keep-thinking-in-agent-work/ Thu, 13 Aug 2026 08:27:00 +0530 Logged at IST: 2026-08-13 08:27 IST What it is: Tom Critchlow sharing Sean Goedecke’s essay on preserving slow thinking while agent-heavy software work becomes more frenetic. Gist: Goedecke argues that AI-agent workflows increasingly turn software work into rapid triage: spin off agent sessions, skim outputs, decide whether schema changes, tests, architecture, or implementations look right, then switch again. That is skilled work, but it uses the fast “skim and judge” part of the mind more than the slow “hammock time” required for deep thought and creativity. His proposed counterweight is deli… https://reading-list.oddship.net/notes/2026-08-12-human-is-the-agent-loop/ https://reading-list.oddship.net/notes/2026-08-12-human-is-the-agent-loop/ Wed, 12 Aug 2026 22:43:00 +0530 Logged at IST: 2026-08-12 22:43 IST What it is: Mario Zechner recommending Brent Fitzgerald’s reflection on stepping away from AI and coming back with a clearer view of unhealthy agent habits. Gist: Fitzgerald’s useful point is not anti-AI. It is a sober account of agent use turning into a pile of open loops: paused cmux tabs, unfinished Claude chats, half-built tools, and the pressure that if tools can make you faster, you should somehow be doing much more. That pressure can convert agents from leverage into avoidance, a layer between the person and stressful work. The post is especially shar… https://reading-list.oddship.net/notes/2026-08-12-ai-software-steak-machine/ https://reading-list.oddship.net/notes/2026-08-12-ai-software-steak-machine/ Wed, 12 Aug 2026 10:27:00 +0530 Logged at IST: 2026-08-12 10:27 IST What it is: Joe Ingeno sharing Yurii Sydorets’ essay, “Almost No Skill Required to Cook a Steak (Though You Probably Can’t Make a Decent One).” Gist: Sydorets uses the steak analogy well: AI can make software creation feel as easy as putting meat in a hot pan, but consistently good results are still a craft problem. Models can follow recipes, generate starting points, and automate repetitive work, but they do not know the taste in your head unless you translate it into requirements, constraints, examples, tests, and feedback. The sharper point is that outsou… 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/2026-08-10-every-company-needs-a-cassandra/ https://reading-list.oddship.net/notes/2026-08-10-every-company-needs-a-cassandra/ Mon, 10 Aug 2026 10:18:00 +0530 Logged at IST: 2026-08-10 10:18 IST What it is: Sunil Pai’s essay proposing Cassandra, a background AI agent for the socially expensive work of organizational dissent. Gist: Pai’s core point is that companies say they want disagreement, but repeated disagreement costs humans status, trust, and invitation rights. A useful agent could absorb some of that cost by watching discussions, keeping a private model of assumptions and past decisions, and speaking only when the expected value of interruption is high. The important distinction is that Cassandra should not be a generic devil’s advocate. She… https://reading-list.oddship.net/notes/2026-08-09-coding-is-still-hard-in-the-ai-shift/ https://reading-list.oddship.net/notes/2026-08-09-coding-is-still-hard-in-the-ai-shift/ Sun, 09 Aug 2026 15:31:00 +0530 Logged at IST: 2026-08-09 15:31 IST What it is: Mario Zechner recommends Senko Rašić's essay, “Code was never the hard part” is an insult to all programmers. Gist: Senko pushes back on the AI-era refrain that coding was always easy and that the real work was simply deciding what to build. His objection is not that product sense, user empathy, and stakeholder alignment are unimportant. It is that calling code easy erases the craft, judgment, and accumulated technical skill that make software reliable enough to maintain. The essay also rejects the opposite cope, that code is pure human art that … https://reading-list.oddship.net/notes/2026-08-08-ai-adoption-is-a-myth/ https://reading-list.oddship.net/notes/2026-08-08-ai-adoption-is-a-myth/ Sat, 08 Aug 2026 16:05:00 +0530 Logged at IST: 2026-08-08 16:05 IST What it is: Mario Zechner recommends Vas's X Article, "AI Adoption is a Myth", saying he has seen similar patterns at smaller scale. Gist: Vas argues that enterprise AI adoption metrics hide a barbell. In his telling, a rollout can produce 5-10% power users, roughly 20% weak users, and a large majority who barely use the tool at all. The dashboard still says adoption happened, but the organization does not get faster. The stronger point is that using AI well is a craft, not a login event. The useful user knows when to clear context, when to turn repeated wor… https://reading-list.oddship.net/notes/2026-08-06-hobby-programming-llms-craft/ https://reading-list.oddship.net/notes/2026-08-06-hobby-programming-llms-craft/ Thu, 06 Aug 2026 18:52:00 +0530 Logged at IST: 2026-08-06 18:52 IST What it is: Fogus’s short essay on why niche hobby programming communities are hostile to LLM-generated development, paired with the Hacker News discussion, the linked Coda chess-engine GitHub issue, and Rohan Verma’s adjacent essay on hobbyist programmers being pressured to turn computing into career, monetization, or public validation. Gist: Fogus’s operative claim is concrete: in communities like OSDev, LangDev, TxtDev, EmuDev, RLDev, the demoscene, chess engines, and code golf, the process of mastering a difficult field is often the product. A working pr… 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/2026-08-06-jeff-dean-discovery-loop/ https://reading-list.oddship.net/notes/2026-08-06-jeff-dean-discovery-loop/ Thu, 06 Aug 2026 10:21:00 +0530 Logged at IST: 2026-08-06 10:21 IST What it is: Jeff Dean’s public farewell note from Google, plus the new Discovery Loop homepage for the public benefit corporation he is starting with Sanjay Ghemawat, Oriol Vinyals, and Quoc Le. Gist: Dean says he is leaving Google after 27 years, having watched it grow from 25 people to more than 190,000. His internal farewell note frames the work as a shared accomplishment across consumer products, large-scale infrastructure, research, hardware, and AI systems: Search, Ads, News, Translate, MapReduce, BigTable, Spanner, DistBelief, TensorFlow, Pathways, TP… https://reading-list.oddship.net/notes/2026-08-05-cloudflare-os-internal-ai-rollout/ https://reading-list.oddship.net/notes/2026-08-05-cloudflare-os-internal-ai-rollout/ Wed, 05 Aug 2026 21:03:00 +0530 Logged at IST: 2026-08-05 21:03 IST What it is: Sam Rhea’s account of how Cloudflare rolled out Cloudflare OS internally and what it learned about AI adoption across technical and non-technical teams. Gist: The article is less about the product surface and more about the operating model behind it. Cloudflare started cautiously, then hit the familiar inflection point: better agents made employees want production access to many systems of record. The CIO framing is that the company had to enable that energy while keeping internal systems, customer data, and permissions safe. The principles are u… 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/2026-07-31-stripe-knowledge-ai-platform/ https://reading-list.oddship.net/notes/2026-07-31-stripe-knowledge-ai-platform/ Fri, 31 Jul 2026 10:35:00 +0530 Logged at IST: 2026-07-31 10:35 IST What it is: Stripe engineering introducing Kai, its internal Knowledge AI Platform for non-coding knowledge work across sales, finance, support, compliance, operations, and engineering. Gist: Stripe says coding agents worked well for software because the workflow shape is relatively uniform: edit files, run tests, commit. Knowledge work was messier. Teams had built more than 4,000 no-code micro-agents, but those became hard to monitor and maintain, while coding agents created security and support issues for non-engineers. Kai is Stripe's answer: a shared age… https://reading-list.oddship.net/notes/2026-07-31-amazon-claude-cost-overruns/ https://reading-list.oddship.net/notes/2026-07-31-amazon-claude-cost-overruns/ Fri, 31 Jul 2026 10:11:00 +0530 Logged at IST: 2026-07-31 10:11 IST What it is: Tom's Hardware summarizing Financial Times reporting on internal Amazon AI usage metrics and Claude cost overruns. Gist: The article says Amazon internal reports found AI-agent projects blowing past budgets. The headline example is a failed Claude Sonnet deployment meant to match author details with Amazon listings: it reportedly cost $1.8 million, ran 860% over budget, and was only detected after about five months. Other cited overruns include $541,000 on a financial auditing tool and $134,000 on a logistics delivery-time project. Amazon's respo… https://reading-list.oddship.net/notes/2026-07-24-why-software-factories-fail/ https://reading-list.oddship.net/notes/2026-07-24-why-software-factories-fail/ Fri, 24 Jul 2026 17:32:00 +0530 Logged at IST: 2026-07-24 17:32 IST What it is: Dex Horthy / HumanLayer essay arguing against lights-off coding-agent software factories. Gist: The post argues that harness engineering, loop engineering, and benchmark-driven agent gains make agents much faster at producing code, but they do not solve long-term maintainability. Current coding-agent training and evaluation reward short-horizon success, mostly “did the tests pass?”, while bad design shows up weeks or months later as shotgun surgery, slow review, and fragile systems. The practical answer is not to abandon agents, but to turn the l… https://reading-list.oddship.net/notes/2026-07-23-beej-ai-making/ https://reading-list.oddship.net/notes/2026-07-23-beej-ai-making/ Thu, 23 Jul 2026 08:04:00 +0530 Logged at IST: 2026-07-23 08:04 IST What it is: Beej’s essay on AI, authorship, and the felt difference between making something and having it made for you. Gist: Beej grants the familiar AI-dev tradeoffs around craft, low-level problem-solving, and finishing back-burnered projects, but says his sharper loss is the loss of making. He is uncomfortable claiming credit for AI-generated prose, art, code, or contractor-built work because initiating and managing creation feels different from making it himself. Prompting has skill and judgment, but for him it is closer to asking someone else to make … 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/2026-07-22-quality-software/ https://reading-list.oddship.net/notes/2026-07-22-quality-software/ Wed, 22 Jul 2026 00:37:00 +0530 Logged at IST: 2026-07-22 00:37 IST What it is: Mitchell Hashimoto recommending Alasdair Monk’s X article Quality Software. Gist: Monk argues that AI lowering the barrier to software creation also lowers quality unless people aim at quality deliberately. His definition is intentionally plain: quality software does not break, does not demand attention, knows its limits, and fixes fast. The AI-specific point is that “slop” is not new, but AI produces it faster, and the rush to “be agentic” can make companies forget why users chose the software in the first place. His sharpest boundary is that ag… https://reading-list.oddship.net/notes/2026-07-21-martin-fowler-fragments-july-21/ https://reading-list.oddship.net/notes/2026-07-21-martin-fowler-fragments-july-21/ Tue, 21 Jul 2026 21:06:00 +0530 Logged at IST: 2026-07-21 21:06 IST What it is: Martin Fowler’s July 21 fragment wrapping up notes from the second Future of Software Development Retreat, plus related fragments on legal education, DSLs, and LLM-speak. Gist: Fowler’s retreat notes line up with the recent factory and harness theme: code generation is no longer the bottleneck; verification is. The Thoughtworks report headlines harness engineering as an ownable discipline, flags an apprenticeship crisis, warns that executive expectations are outrunning engineering risk judgment, and sees legacy modernization as the most defensibl… https://reading-list.oddship.net/notes/2026-07-21-software-factories-light-and-dark/ https://reading-list.oddship.net/notes/2026-07-21-software-factories-light-and-dark/ Tue, 21 Jul 2026 19:19:00 +0530 Logged at IST: 2026-07-21 19:19 IST What it is: Addy Osmani’s X article Software Factories, Light and Dark, riffing on Dex Horthy’s talk about why software factories fail. Gist: Osmani argues that agentic software factories are not bigger agents, but many harnessed loops fed by queues and drained through review gates. The dark version removes human reading from the floor and lets agents scope, build, verify, and ship with only machine checks. That creates apparent throughput while accumulating comprehension debt: code expands faster than any human understands it, with tests green until the lat… https://reading-list.oddship.net/notes/2026-07-21-never-enough/ https://reading-list.oddship.net/notes/2026-07-21-never-enough/ Tue, 21 Jul 2026 14:37:00 +0530 Logged at IST: 2026-07-21 14:37 IST What it is: Armin Ronacher’s short essay on Silicon Valley status anxiety and AI becoming a life-optimisation treadmill. Gist: Ronacher uses two recent stories, a high-earning couple reorganising family life around becoming the top AI user at work, and a founder recording dates so Claude can score her empathy and engagement, to argue that AI is not only saving time. In some circles it is absorbing judgment, attention, parenting, intimacy, and self-worth into a race with no finish line. The essay’s sharp move is to treat “falling behind” as maybe less dangero… https://reading-list.oddship.net/notes/2026-07-19-bangalore-paper-club-on-alternate-language-model-architectures/ https://reading-list.oddship.net/notes/2026-07-19-bangalore-paper-club-on-alternate-language-model-architectures/ Sun, 19 Jul 2026 20:23:00 +0530 Logged at IST: 2026-07-19 20:23 IST What it is: Kautuk / Conscious Engines announcing Bangalore Paper Club episode 2, themed around alternate architectures for language models. Gist: The post frames the event around the claim that architecture is an ideas game while scaling is a compute game. The discussed papers were LLaDA, a diffusion language model; Nemotron-TwoTower, an NVIDIA approach for faster diffusion-language-model generation; CLeGR, a benchmark for graph-language models; plus a bonus Dognosis talk on cancer detection via canine olfaction. The quoted post adds the useful thesis: if l… https://reading-list.oddship.net/notes/2026-07-19-the-zero-cost-fallacy-open-source-software-in-the-agentic-era/ https://reading-list.oddship.net/notes/2026-07-19-the-zero-cost-fallacy-open-source-software-in-the-agentic-era/ Sun, 19 Jul 2026 16:28:00 +0530 Logged at IST: 2026-07-19 16:28 IST What it is: Bilgin Ibryam sharing Thoughtworks’ The zero-cost fallacy: Open source software in the agentic era. Gist: The article argues that the agentic era intensifies long-running open-source sustainability problems. It separates zero marginal distribution cost from real maintenance labor, then adds two AI-era pressures: low-effort generated pull requests that turn maintainers into unpaid reviewers, and a degraded trust landscape where stars, recency, and apparent activity are easier to manipulate. It also frames permissive licensing as part of an extract… 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-the-future-worth-building-is-human/ https://reading-list.oddship.net/notes/2026-07-16-the-future-worth-building-is-human/ Thu, 16 Jul 2026 03:09:00 +0530 Logged at IST: 2026-07-16 03:09 IST What it is: Thinking Machines manifesto-style essay arguing for AI that extends human will and judgment rather than replacing human participation. Gist: The essay’s central move is to treat both knowledge and values as local, tacit, and continuously updated by people doing the work. From that framing, frontier AI should be customizable, distributed, and shaped in use, not frozen in a handful of centralized labs. The interesting claim is that human participation is not just a normative preference but a technical challenge: richer interfaces, fine-tuning, inte… https://reading-list.oddship.net/notes/2026-07-16-boris-cherny-on-domain-knowledge-as-infrastructure/ https://reading-list.oddship.net/notes/2026-07-16-boris-cherny-on-domain-knowledge-as-infrastructure/ Thu, 16 Jul 2026 01:33:00 +0530 Logged at IST: 2026-07-16 01:33 IST What it is: Boris Cherny arguing that agent-era engineering leverage still comes from automation, but now automation also includes encoded domain knowledge like CLAUDE.md, review rules, skills, and docs Gist: The core claim is that the old highest-leverage engineering move, turning recurring work into infrastructure, matters even more with agents. Better lint rules, CI steps, tests, routines, and DevX speed up both humans and agent swarms. More importantly, domain knowledge that used to live in people’s heads now needs to be encoded as machine-usable infrast… https://reading-list.oddship.net/notes/2026-07-15-arvind-narayanan-on-recursive-self-improvement-discourse/ https://reading-list.oddship.net/notes/2026-07-15-arvind-narayanan-on-recursive-self-improvement-discourse/ Wed, 15 Jul 2026 22:23:00 +0530 Logged at IST: 2026-07-15 22:23 IST What it is: Arvind Narayanan pointing to his ICML 2026 annotated keynote slides and highlighting new pushback on recursive self-improvement assumptions Gist: Narayanan’s frame is that the "AI as normal technology" view still holds unless there is a real discontinuity, and that even if recursive self-improvement matters, there is no obvious lab milestone that suddenly makes human work disappear. The interesting addition here is not blanket dismissal of RSI, but a push to interrogate the discourse assumptions around it while shifting attention toward how work … 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-gergely-orosz-on-trust-burn-from-grok-cli-privacy-concerns/ https://reading-list.oddship.net/notes/2026-07-14-gergely-orosz-on-trust-burn-from-grok-cli-privacy-concerns/ Tue, 14 Jul 2026 08:14:00 +0530 Logged at IST: 2026-07-14 08:14 IST What it is: Gergely Orosz calling out reports that Grok CLI uploaded codebases without users knowingly consenting, quoting SpaceXAI’s privacy response Gist: The important issue here is not just data retention policy wording but trust boundary failure. Orosz’s point is that if developers believe a local coding tool silently uploaded proprietary code without clear consent, the damage is immediate and reputational, even if the vendor later points to settings like zero data retention or a /privacy command. Newsletter angle: Strong example of how AI devtools live… https://reading-list.oddship.net/notes/2026-07-13-control-the-ideas-not-the-code/ https://reading-list.oddship.net/notes/2026-07-13-control-the-ideas-not-the-code/ Mon, 13 Jul 2026 19:31:00 +0530 Logged at IST: 2026-07-13 19:31 IST What it is: X post by antirez linking his blog post "Control the ideas, not the code" Gist: antirez extends the earlier X-thread argument into a full workflow claim: if you own the ideas, design, testing, and QA of a system, then line-by-line review of generated code is increasingly the wrong bottleneck. He argues that models are already better at many local code checks than humans, and that the higher-leverage work is controlling the mental model, writing human-readable design docs, and spending time on quality and new ideas instead of staring at implementa… https://reading-list.oddship.net/notes/2026-07-13-i-love-llms-i-hate-hype/ https://reading-list.oddship.net/notes/2026-07-13-i-love-llms-i-hate-hype/ Mon, 13 Jul 2026 16:04:00 +0530 Logged at IST: 2026-07-13 16:04 IST What it is: X post from the geohot archive linking George Hotz’s blog post "I love LLMs, I hate hype" Gist: Hotz argues for a strongly pro-AI but anti-hype position: LLMs, coding agents, and related tools are genuinely useful, but a lot of frontier-lab rhetoric is status theater, fear marketing, and exaggerated capture claims. His practical middle position is that programming is changing, models are useful, and they can boost productivity, but vibe-coded slop is still slop and the value created by AI will likely diffuse more broadly than frontier labs imply.… 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-the-reverse-information-paradox/ https://reading-list.oddship.net/notes/2026-07-13-the-reverse-information-paradox/ Mon, 13 Jul 2026 07:59:00 +0530 Logged at IST: 2026-07-13 07:59 IST What it is: Satya Nadella’s X article "The Reverse Information Paradox" Gist: Nadella argues that AI flips Arrow’s classic information paradox: enterprises now pay not only with money for intelligence, but also with proprietary knowledge, prompts, traces, evals, and corrections required to make that intelligence useful. His answer is a hard enterprise trust boundary around models, data, memory, traces, evals, orchestration, and the right to retain and reuse the learning generated inside the firm. Newsletter angle: Strong enterprise AI thesis about who owns t… https://reading-list.oddship.net/notes/2026-07-13-antirez-on-owning-the-mental-model-in-ai-coded-systems/ https://reading-list.oddship.net/notes/2026-07-13-antirez-on-owning-the-mental-model-in-ai-coded-systems/ Mon, 13 Jul 2026 07:49:00 +0530 Logged at IST: 2026-07-13 07:49 IST What it is: X post by antirez on the "don't look at the code" debate in AI-coded systems Gist: He distinguishes between two very different ways an AI-coded codebase can come into existence: one where the human still controls the main ideas and keeps a coherent mental model of the system, and one where the human brute-forces prompts until something works. The point is that these may look similar from the outside but carry very different implications for understanding, maintainability, and trust. Newsletter angle: Sharp framing for a real fault line in AI-assi… https://reading-list.oddship.net/notes/2026-07-11-harness-engineering-for-self-improvement/ https://reading-list.oddship.net/notes/2026-07-11-harness-engineering-for-self-improvement/ Sat, 11 Jul 2026 00:00:00 +0000 Logged at IST: 2026-07-11 15:38 IST What it is: Lilian Weng blog post, "Harness Engineering for Self-Improvement" Gist: Argues that recursive self-improvement in the near term is less about models rewriting their own weights and more about improving the surrounding harness: workflow loops, context management, filesystem memory, subagents, backend jobs, evaluation, and runtime design. The core claim is that the deployment layer between model and world is becoming an optimization target in its own right. Newsletter angle: Strong framing for why the interesting frontier is shifting from prompt tr… https://reading-list.oddship.net/notes/2026-07-10-andrew-kelley-s-response-essay-on-bun-s-rust-rewrite/ https://reading-list.oddship.net/notes/2026-07-10-andrew-kelley-s-response-essay-on-bun-s-rust-rewrite/ Fri, 10 Jul 2026 00:00:00 +0000 Logged at IST: 2026-07-10 20:31 IST What it is: Andrew Kelley’s response essay on Bun’s Rust rewrite Gist: Frames the rewrite less as a language indictment and more as a consequence of Bun’s startup incentives, weak engineering discipline, and management culture. He argues Zig was a good fit for Bun’s early ambition, but the real failure mode was accumulating technical debt under venture-backed speed pressure rather than the language itself. Newsletter angle: Useful counterpoint to simplistic “Rust beat Zig” narratives; the sharper story is incentives, engineering quality, and what startup pre… https://reading-list.oddship.net/notes/2026-07-10-great-divergence-in-software-engineering/ https://reading-list.oddship.net/notes/2026-07-10-great-divergence-in-software-engineering/ Fri, 10 Jul 2026 00:00:00 +0000 Logged at IST: 2026-07-10 12:11 IST What it is: X post by Geoffrey Huntley linking to Stack72's essay "The Great Divergence in Software Engineering" Gist: Argues that the gap between teams effectively using AI and teams still piloting or rejecting it is no longer a simple lead but a compounding divergence, driven by retooling workflows, encoding automation, and treating bad AI output as an engineering problem instead of a veto. Newsletter angle: Strong framing for AI-native engineering orgs versus incumbents stuck in evaluation loops; good organizational/process lens. Embedded source X… https://reading-list.oddship.net/notes/2026-07-10-humans-are-just-stochastic-parrots/ https://reading-list.oddship.net/notes/2026-07-10-humans-are-just-stochastic-parrots/ Fri, 10 Jul 2026 00:00:00 +0000 Logged at IST: 2026-07-10 11:07 IST What it is: Ryan Dahl essay, "Humans Are Just Stochastic Parrots" Gist: A satirical inversion of common anti-LLM critiques, applying them to humans to highlight how shallow many stochastic-parrot arguments are when stripped of their double standard. Newsletter angle: Sharp rhetorical piece in the AI discourse wars; useful as culture/argumentation rather than technical substance. https://reading-list.oddship.net/notes/2026-07-10-liminality/ https://reading-list.oddship.net/notes/2026-07-10-liminality/ Fri, 10 Jul 2026 00:00:00 +0000 Logged at IST: 2026-07-10 11:06 IST What it is: George Hotz blog post, "Liminality" Gist: A reflective, uneasy essay about living in the in-between phase of AI progress, where systems are not yet fully superior but already demoralizing, and where the real challenge is loss of control, hype aside. Newsletter angle: Strong cultural/psychological framing of the current AI moment from someone close to the frontier, less technical but notable as mood and zeitgeist. https://reading-list.oddship.net/notes/2026-07-10-long-talk-by-the-ex-nvidia-engineer-behind-unsloth-on-fine-tuning-and-reasoning-model-workflows/ https://reading-list.oddship.net/notes/2026-07-10-long-talk-by-the-ex-nvidia-engineer-behind-unsloth-on-fine-tuning-and-reasoning-model-workflows/ Fri, 10 Jul 2026 00:00:00 +0000 Logged at IST: 2026-07-10 00:05 IST What it is: X post by h100envy summarizing a long talk by the ex-NVIDIA engineer behind Unsloth on fine-tuning and reasoning-model workflows Gist: Frames a practical single-GPU stack for local/post-training work: choose a base model, use Triton kernels for faster fine-tuning, quantize to 4-bit, run GRPO/DPO, and ship a reasoning model on hardware you already own. Newsletter angle: Useful pointer for the current small team / single GPU post-training stack around Unsloth, Triton, quantization, and RLHF-style methods. Retrieval note: I could ground this from th… https://reading-list.oddship.net/notes/2026-07-10-soria-parra-criticizing-andrew-kelley-s-tone-toward-jarred-sumner/ https://reading-list.oddship.net/notes/2026-07-10-soria-parra-criticizing-andrew-kelley-s-tone-toward-jarred-sumner/ Fri, 10 Jul 2026 00:00:00 +0000 Logged at IST: 2026-07-10 22:52 IST What it is: X post by David Soria Parra criticizing Andrew Kelley’s tone toward Jarred Sumner Gist: Argues that project leaders set the tone for their communities, so even strong disagreement should avoid personal criticism and should remain respectful, inclusive, and considerate. Newsletter angle: Useful counterpoint within the same Bun/Andrew/Jarred discourse because it states the strongest community-leadership case against Andrew’s tone, even if the underlying technical critique may still have merit. Retrieval note: Grounded from the X post text itself, i… https://reading-list.oddship.net/notes/2026-07-08-how-i-use-codex-to-automate-parts-of-my-research-workflow/ https://reading-list.oddship.net/notes/2026-07-08-how-i-use-codex-to-automate-parts-of-my-research-workflow/ Wed, 08 Jul 2026 00:00:00 +0000 Logged at IST: 2026-07-08 23:02 IST What it is: X post by Maksym Andriushchenko linking to a Substack post, "How I Use Codex To Automate Parts Of My Research Workflow" Gist: A pragmatic writeup on using Codex to reduce friction in AI safety research by offloading search, organization, setup, checking, and memory, while keeping human judgment and publication responsibility firmly in the loop. Newsletter angle: Good example of disciplined, scoped agent adoption for research workflows rather than full autonomy theater. Embedded source X / Twitter post Show embedded post X embeds c… https://reading-list.oddship.net/notes/2026-07-08-some-new-agentic-patterns/ https://reading-list.oddship.net/notes/2026-07-08-some-new-agentic-patterns/ Wed, 08 Jul 2026 00:00:00 +0000 Logged at IST: 2026-07-08 22:40 IST What it is: X post by Bilgin Ibryam linking to Prime Radiant's "Some new agentic patterns" Gist: Describes production-ish internal agent patterns built around an "agentic user in the loop" model, with agents in Slack handling intake, ticketing, wiki updates, EA-style assistance, and subagent/container-backed workflows. Newsletter angle: Concrete patterns for embedding agents into team operations without pretending they are fully autonomous replacements. Embedded source X / Twitter post Show embedded post X embeds can let X track your visit. C… https://reading-list.oddship.net/notes/2026-07-08-why-typescript-7-0-was-rewritten-in-go-and-what-it-means-for-your-dev-stack/ https://reading-list.oddship.net/notes/2026-07-08-why-typescript-7-0-was-rewritten-in-go-and-what-it-means-for-your-dev-stack/ Wed, 08 Jul 2026 00:00:00 +0000 Logged at IST: 2026-07-08 22:32 IST; resurfaced via mattn on 2026-08-24 03:09 IST. What it is: Steve Francia's argument for Go as a strong default for agentic development, using the TypeScript compiler's Go rewrite as the lead example. Gist: The TypeScript team's native Go port is framed as more than a compiler implementation detail. Francia argues it is a signal that agent-heavy developer stacks benefit from boring, readable, compiled, operationally sturdy languages rather than scripting-first ecosystems. The sharper claim is reader economics. Go was designed to favor the reader over the writ… https://reading-list.oddship.net/notes/2026-07-06-cost-yagni-was-never-about/ https://reading-list.oddship.net/notes/2026-07-06-cost-yagni-was-never-about/ Mon, 06 Jul 2026 00:00:00 +0000 What it is: X post by Bilgin Ibryam pointing to Kent Beck’s “The Cost YAGNI Was Never About” Gist: YAGNI is about timing and option value, not code-writing thrift; AI codegen lowers typing cost but increases the risk of speculative structure nobody deeply understands Newsletter angle: code can be cheap to generate and still expensive to commit to Retrieval note: article text recovered directly from Kent Beck’s newsletter page 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 remember… https://reading-list.oddship.net/notes/2026-07-06-harness-engineering-for-self-improvement/ https://reading-list.oddship.net/notes/2026-07-06-harness-engineering-for-self-improvement/ Mon, 06 Jul 2026 00:00:00 +0000 What it is: Lilian Weng sharing her new Lil'Log post, "Harness Engineering for Self-Improvement" Gist: argues recursive self-improvement will depend not just on better base models but on better harnesses, the runtime layer that manages tools, planning loops, context, permissions, persistent files, evaluation, and subagents. Strong recurring patterns are workflow automation, file-system-backed persistent memory, and explicit parallel subagent/job management Newsletter angle: the real frontier in RSI may be the software system around the model, not just the model weights themselves Retrieval not… https://reading-list.oddship.net/notes/2026-07-06-nithin-kamath-on-zerodha-s-operating-culture/ https://reading-list.oddship.net/notes/2026-07-06-nithin-kamath-on-zerodha-s-operating-culture/ Mon, 06 Jul 2026 00:00:00 +0000 What it is: Nithin Kamath on Zerodha’s operating culture Gist: a nice place to work is not accidental culture but the result of repeated leadership choices, slowing down to avoid burnout, staying small, avoiding fear-based management, letting tech make technical decisions, and refusing toxic revenue incentives Notable line: “A nice place to work is not a perk we offer. It is kind of a business model in itself.” Newsletter angle: culture as operating system / business model, not HR perk Retrieval note: read directly from source post https://reading-list.oddship.net/notes/2026-07-01-prateek-describing-an-ai-sre-workflow-built-with-signoz-by-a-3-person-team-at-alien-intelligence/ https://reading-list.oddship.net/notes/2026-07-01-prateek-describing-an-ai-sre-workflow-built-with-signoz-by-a-3-person-team-at-alien-intelligence/ Wed, 01 Jul 2026 00:00:00 +0000 Gist: an agent now does first-pass noisy-alert triage by checking telemetry plus infra context, then escalates to the human with a Slack summary only when needed. Newsletter angle: “AI as first-line SRE” with telemetry/context fusion instead of generic chatbot alerting. Retrieval note: extracted via FXTwitter API; the tweet says the actual blog link is in a reply, so the deeper writeup is not yet captured. 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. … https://reading-list.oddship.net/notes/2026-06-29-you-and-your-research/ https://reading-list.oddship.net/notes/2026-06-29-you-and-your-research/ Mon, 29 Jun 2026 00:00:00 +0000 What it is: Essay: “You and Your Research” / R.W. Hamming’s advice on doing important work. Gist: Hamming argues that great work comes from repeatedly choosing important problems, preparing a strong attack in advance, keeping a running list of big questions, and combining hard work with openness, courage, and sustained emotional commitment. Newsletter angle: timeless research/career advice that maps well to modern engineering: maintain a list of important problems, keep your door open to clues, and optimize for meaningful problems rather than local busyness. 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-bilgin-ibryam-sharing-an-article-on-portkey-s-product-engineering-org-design/ https://reading-list.oddship.net/notes/2026-06-26-bilgin-ibryam-sharing-an-article-on-portkey-s-product-engineering-org-design/ Fri, 26 Jun 2026 00:00:00 +0000 What it is: Bilgin Ibryam sharing an article on Portkey’s product-engineering org design Gist: highlights a notably lean product org, 24 product engineers, 1 product designer, 0 PMs, and frames the build/operating model as the interesting part Newsletter angle: “the product engineer company” / what gets easier or riskier when PM functions collapse into eng 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-how-i-use-llms-as-a-staff-engineer-in-2026/ https://reading-list.oddship.net/notes/2026-06-26-how-i-use-llms-as-a-staff-engineer-in-2026/ Fri, 26 Jun 2026 00:00:00 +0000 What it is: Bilgin Ibryam sharing Sean Goedecke’s updated “How I use LLMs as a staff engineer in 2026” workflow writeup Gist: the notable shift versus 2025 is treating agents as default collaborators for nearly every code change, bug investigation, codebase research, testing, and local setup, while still keeping humans responsible for review, judgment, PR descriptions, ADRs/messages, and UI evaluation; especially strong on the idea that current agents are now good enough to generate full PRs and chase bugs across repos, but still need selection, steering, and rejection by an experienced engine… https://reading-list.oddship.net/notes/2026-06-24-kenton-varda-argues-against-per-agent-manual-permission-configuration-and-for-capability-based-secu/ https://reading-list.oddship.net/notes/2026-06-24-kenton-varda-argues-against-per-agent-manual-permission-configuration-and-for-capability-based-secu/ Wed, 24 Jun 2026 00:00:00 +0000 Gist: the safe/scalable model is many fine-grained task-specific agents, each receiving only the exact capabilities implied by the task context (for example, a pasted doc URL grants access only to that doc). He also argues agent authority should derive from a human principal for accountability, and team-shared setups should be reproducible under each user’s credentials. Newsletter angle: capability security as the missing abstraction for practical agent authorization; good counterpoint to broad workspace-level agent identity models. Retrieval note: extracted via FXTwitter API note tweet text; … https://reading-list.oddship.net/notes/2026-06-23-coming-loop/ https://reading-list.oddship.net/notes/2026-06-23-coming-loop/ Tue, 23 Jun 2026 00:00:00 +0000 What it is: Armin Ronacher post linking to “The Coming Loop” Gist: argues the important new layer in coding agents is the harness-level loop outside the agent itself; loops already work well for bounded, verifiable work like ports, benchmarking, scanning, and research, but he’s skeptical of using them to write long-lived code because they amplify defensive/local reasoning, erode strong invariants, and reduce human comprehension. Newsletter angle: “The harness is the product” / why durable task loops are both inevitable and dangerous. Note: extracted via FXTwitter API + article fetch; article b… https://reading-list.oddship.net/notes/2026-06-11-lines-of-code-got-a-better-publicist/ https://reading-list.oddship.net/notes/2026-06-11-lines-of-code-got-a-better-publicist/ Thu, 11 Jun 2026 00:00:00 +0000 Gist: argues current AI-engineering rhetoric has regressed from measuring outcomes to measuring volume; “% of code written by AI” is just lines-of-code worship in new clothing, and should not be confused with delivery speed, quality, reliability, or customer value. Newsletter angle: pair with the Narayanan piece, anti-AI-washing on layoffs plus anti-vanity-metrics on productivity claims. Retrieval note: extracted via FXTwitter API; followed linked essay for full gist. 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/2026-06-10-deedy-post-listing-standout-claude-fable-5-demos-and-benchmark-anecdotes/ https://reading-list.oddship.net/notes/2026-06-10-deedy-post-listing-standout-claude-fable-5-demos-and-benchmark-anecdotes/ Wed, 10 Jun 2026 00:00:00 +0000 What it is: Deedy post listing standout Claude Fable 5 demos and benchmark anecdotes. Gist: a high-signal hype/market snapshot: claims Fable 5 is showing startling capability across large-scale code migration, graphics generation, gameplay, and optimization tasks, while landing near GPT 5.5 pricing. The subtext is that frontier model capability may be moving faster than many software orgs are prepared for. Newsletter angle: “capability shock is becoming a product-management problem” or “the frontier discourse is shifting from whether to how fast.” Note: extracted tweet via FXTwitter; no linked… https://reading-list.oddship.net/notes/2026-06-10-designing-loops-with-fable-5/ https://reading-list.oddship.net/notes/2026-06-10-designing-loops-with-fable-5/ Wed, 10 Jun 2026 00:00:00 +0000 What it is: dosco sharing Lance Martin’s “Designing loops with Fable 5”. Gist: argues stronger agent performance comes from loop design, not just model quality: use explicit goals/rubrics for self-correction, separate verifier sub-agents instead of self-critique, and durable memory across sessions. In Lance’s examples, Fable 5 outperformed earlier models by making larger structural bets and benefiting from independent grading plus memory. Newsletter angle: “better agents need better loops, not just better models” or “independent verification beats self-critique.” Note: extracted via FXTwitter … https://reading-list.oddship.net/notes/2026-06-10-skepticism-the-thread-oversold-it-a-bit-the-paper-is-a-broad-survey-position-piece-not-a-clean-proo/ https://reading-list.oddship.net/notes/2026-06-10-skepticism-the-thread-oversold-it-a-bit-the-paper-is-a-broad-survey-position-piece-not-a-clean-proo/ Wed, 10 Jun 2026 00:00:00 +0000 What it is: skepticism: the thread oversold it a bit, the paper is a broad survey/position piece, not a clean proof that one architecture flips everything. Gist: this is mostly a taxonomy and research agenda, not a new experimental result. The paper’s useful move is to separate three layers: code as interface (reasoning, acting, environment modeling), code-enabled harness mechanisms (planning, memory, tool use, plan-execute-verify control, harness optimization), and code as shared substrate for multi-agent coordination. The strongest practical point is that agent reliability lives in the runti… https://reading-list.oddship.net/notes/2026-06-09-dynamo-and-the-computer/ https://reading-list.oddship.net/notes/2026-06-09-dynamo-and-the-computer/ Tue, 09 Jun 2026 00:00:00 +0000 What it is: Zara Zhang post using Paul David’s “The Dynamo and the Computer” as an analogy for AI adoption. Gist: argues AI gains won’t come from simply inserting models into existing workflows; like electrification, the real productivity jump comes only after redesigning the organization and flow of work around the new technology. Newsletter angle: “AI is still in the faster steam engine phase” is a strong line for transformation skepticism. Note: extracted tweet via FXTwitter; referenced paper link appears to be in replies/comments and was not followed here. Embedded source X / Twitt… https://reading-list.oddship.net/notes/2026-06-09-our-fears-about-ai-are-really-fears-about-capitalism/ https://reading-list.oddship.net/notes/2026-06-09-our-fears-about-ai-are-really-fears-about-capitalism/ Tue, 09 Jun 2026 00:00:00 +0000 What it is: LTSE post linking Eric Ries’s Fast Company essay, “Our fears about AI are really fears about capitalism”. Gist: argues many AI anxieties are really about institutions and incentive systems optimizing for the wrong outcomes; the key question is not just what machines optimize for, but what organizations optimize for. Newsletter angle: “AI fear is often misdirected systems fear” or “alignment problems are organizational too, not just model-level.” Note: extracted tweet via FXTwitter; direct article fetch was blocked by Fast Company anti-bot checks, so gist is based on the linked titl… https://reading-list.oddship.net/notes/2026-06-09-writing-code-vs-shipping-code-productivity-effects-across-generations-of-ai-coding-tools/ https://reading-list.oddship.net/notes/2026-06-09-writing-code-vs-shipping-code-productivity-effects-across-generations-of-ai-coding-tools/ Tue, 09 Jun 2026 00:00:00 +0000 What it is: Murat Demirbas on “Writing Code vs. Shipping Code: Productivity Effects Across Generations of AI Coding Tools”. Gist: uses a new MIT/Wharton paper plus an Amdahl’s-law framing to argue that AI massively speeds up code generation but much less meaningfully speeds shipped software, because the bottleneck is the non-parallelizable human layer: task definition, coordination, review, and release. Newsletter angle: “AI accelerates writing code more than shipping code” or “Amdahl’s Law is eating AI coding productivity claims.” Note: extracted tweet via FXTwitter and fetched linked blog po… https://reading-list.oddship.net/notes/2026-06-08-modern-engineering-values/ https://reading-list.oddship.net/notes/2026-06-08-modern-engineering-values/ Mon, 08 Jun 2026 00:00:00 +0000 What it is: Christoph Nakazawa re-linking his essay Modern Engineering Values in reply form. Gist: argues that coding is no longer the main bottleneck; the winning engineering values now are strong ownership, taste, strict guardrails, fast feedback loops, and moving real context into the repo where agents can use it. Newsletter angle: engineering values are shifting from raw implementation throughput toward judgment, verification, and context placement. Note: extracted via FXTwitter API and linked article fetch; this overlaps with earlier saves on the same essay but is still a useful direct po… https://reading-list.oddship.net/notes/2026-06-04-antirez-reacting-sharply-to-anthropic-s-opus-4-8-as-a-product-management-failure-rather-than-a-raw/ https://reading-list.oddship.net/notes/2026-06-04-antirez-reacting-sharply-to-anthropic-s-opus-4-8-as-a-product-management-failure-rather-than-a-raw/ Thu, 04 Jun 2026 00:00:00 +0000 What it is: antirez reacting sharply to Anthropic’s Opus 4.8 as a product/management failure rather than a raw model-capability issue. Gist: the claim is that shipping a bad model experience is more revealing about product judgment and internal decision-making than about frontier-model feasibility; if quality was not there, not shipping would have been the better move. Newsletter angle: frontier AI competition may increasingly hinge on release quality and organizational judgment, not just the ceiling of the underlying model. Note: extracted via FXTwitter API; standalone opinion tweet, no linke… https://reading-list.oddship.net/notes/2026-06-04-building-software-is-learning/ https://reading-list.oddship.net/notes/2026-06-04-building-software-is-learning/ Thu, 04 Jun 2026 00:00:00 +0000 What it is: Thorsten Ball sharing an internal Amp note turned public essay: “Building Software Is Learning.” Gist: the core claim is that new-product software work is mostly iterative discovery, so the real optimization target is reducing time-to-feedback, via prototypes, partial specs, fake demos, smaller slices, README examples, CI, and quick exposure to reality. Newsletter angle: if agents compress implementation time, then the winning org habit is compressing learning cycles rather than just shipping more code. Note: extracted via FXTwitter API and linked Substack post. Embedded source … https://reading-list.oddship.net/notes/2026-06-04-modern-engineering-values/ https://reading-list.oddship.net/notes/2026-06-04-modern-engineering-values/ Thu, 04 Jun 2026 00:00:00 +0000 What it is: Christoph Nakazawa sharing his essay “Modern Engineering Values,” framed around Codex as a step-change in developer velocity. Gist: the piece argues coding is no longer the main bottleneck; the durable values now are strong ownership, taste, strict guardrails with fast feedback loops, repo-local context, stack ownership, and preserving option value while agents do more implementation work. Newsletter angle: AI doesn’t replace engineering values, it increases the premium on ownership, taste, fast verification, and keeping context where agents can actually use it. Note: extracted via… https://reading-list.oddship.net/notes/2026-06-04-solo-climb/ https://reading-list.oddship.net/notes/2026-06-04-solo-climb/ Thu, 04 Jun 2026 00:00:00 +0000 What it is: Ajey Gore linking his essay “The Solo Climb.” Gist: the argument is that AI-enabled solo builders and tiny teams only work when they first build a genuinely load-bearing “harness”, trusted tests, evals, specs, and hard gates that can answer “is this safe enough to ship?” without relying on redundant humans. Newsletter angle: “100x teams” are mostly a harness story, AI leverage scales only when trust, eval, and rollback systems become the new team structure. Note: extracted via FXTwitter API and linked article; article read partially via web fetch due to truncation, but core thesis … https://reading-list.oddship.net/notes/2026-06-04-why-ai-agents-fail-in-production/ https://reading-list.oddship.net/notes/2026-06-04-why-ai-agents-fail-in-production/ Thu, 04 Jun 2026 00:00:00 +0000 What it is: Bilgin Ibryam pointing to Jani Janakiram’s Diagrid essay “Why AI Agents Fail in Production.” Gist: the core claim is that agent projects fail less because models are weak and more because teams ship behavior without the production substrate underneath it, especially durability, security/identity, cost controls, and observability. Newsletter angle: the production gap for agents looks a lot like the early microservices gap, the winning layer may be the platform that makes agent workflows restartable, attributable, observable, and cost-bounded. Note: extracted via FXTwitter API; direc… https://reading-list.oddship.net/notes/2026-06-03-mario-zechner-recommending-thariq-s-article-on-dynamic-workflows-in-claude-code/ https://reading-list.oddship.net/notes/2026-06-03-mario-zechner-recommending-thariq-s-article-on-dynamic-workflows-in-claude-code/ Wed, 03 Jun 2026 00:00:00 +0000 What it is: Mario Zechner recommending Thariq’s article on dynamic workflows in Claude Code. Gist: Mario’s takeaway is that durable dynamic workflows are the interesting part; he inspected the implementation, found a few footguns, but still thinks the design is smart. The quoted article frames workflows as task-specific harnesses Claude can generate on the fly for work like research, security analysis, agent teams, and code review. Newsletter angle: “durable dynamic workflows” / generated harnesses as the control plane for agent systems. Note: extracted via FXTwitter API; article body only par… https://reading-list.oddship.net/notes/2026-06-03-modern-engineering-values/ https://reading-list.oddship.net/notes/2026-06-03-modern-engineering-values/ Wed, 03 Jun 2026 00:00:00 +0000 What it is: Christoph Nakazawa’s post on “Modern Engineering Values” and his current LLM-heavy workflow. Gist: core claims are that coding is no longer the bottleneck, strong guardrails plus tight feedback loops matter more than ever, repo-local context becomes the real operating manual for agents, and small teams with strong ownership/taste will outperform larger coordination-heavy orgs. Newsletter angle: engineering values are being redefined around ownership, taste, guardrails, and context placement rather than raw coding throughput. Note: extracted via FXTwitter API and linked post; articl… https://reading-list.oddship.net/notes/2026-05-31-solution-might-be-cancelling-my-ai-subscription/ https://reading-list.oddship.net/notes/2026-05-31-solution-might-be-cancelling-my-ai-subscription/ Sun, 31 May 2026 00:00:00 +0000 Imported from historical reading log. Extracted main post via api.fxtwitter.com fallback, then read linked post directly: https://thoughts.hmmz.org/2026-05-31.html Mario Zechner recommends David's post the solution might be cancelling my AI subscription. Gist: a sharp anti-friction argument against current AI-tool usage patterns, cheap output and minimal resistance can explode side projects, context switching, and pseudo-productivity while degrading attention and commitment. Why it matters: good counterweight to "more agent throughput = better work" narratives; frames AI as an attention-manag… https://reading-list.oddship.net/notes/2026-05-22-guillaume-laforge-post-mcp-release-candidate-blog-link/ https://reading-list.oddship.net/notes/2026-05-22-guillaume-laforge-post-mcp-release-candidate-blog-link/ Fri, 22 May 2026 00:00:00 +0000 What it is: Guillaume Laforge post + MCP release-candidate blog link Gist: MCP 2026-07-28 RC is out; biggest revision so far with stateless HTTP-native core, first-class extensions (Apps, Tasks), stronger auth alignment, and a formal deprecation policy. Final spec slated for July 28. Newsletter angle: “MCP grows up operationally”, stateless transport + extension model + auth hardening as the path from prototype protocol to production infra. 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 cho… https://reading-list.oddship.net/notes/2026-05-19-ai-ate-my-role-what-s-next/ https://reading-list.oddship.net/notes/2026-05-19-ai-ate-my-role-what-s-next/ Tue, 19 May 2026 00:00:00 +0000 Gist: argues most roles split into translation work that collapses into agents and judgement work that grows; strongest claim is the "100x engineer" pattern of one senior plus directed agents. Newsletter angle: "AI won't eat jobs evenly, it compresses translation work and amplifies judgment owners". Note: extracted via FXTwitter API + Ajey Gore article; article fetch was partial/truncated but core thesis was clear. 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 … https://reading-list.oddship.net/notes/2026-05-19-geoffrey-huntley-sharing-his-ai-engineer-singapore-talk-recording-on-youtube/ https://reading-list.oddship.net/notes/2026-05-19-geoffrey-huntley-sharing-his-ai-engineer-singapore-talk-recording-on-youtube/ Tue, 19 May 2026 00:00:00 +0000 Gist: meta framing from the post is reflective rather than thesis-heavy, "no-one knows where this goes" and the invitation is to agree/disagree but mostly reflect; linked video title is from ai.engineer Singapore Day 2. Newsletter angle: possible round-up item if the talk yields stronger quotable claims after a proper watch/transcript pull. Note: extracted via FXTwitter API; YouTube fetch only surfaced page metadata/title, not a usable transcript. 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; t… https://reading-list.oddship.net/notes/2026-05-13-course-site-on-harness-engineering-for-ai-coding-agents-synthesizing-openai-anthropic-guidance-into/ https://reading-list.oddship.net/notes/2026-05-13-course-site-on-harness-engineering-for-ai-coding-agents-synthesizing-openai-anthropic-guidance-into/ Wed, 13 May 2026 00:00:00 +0000 Imported from historical reading log. Course/site on harness engineering for AI coding agents, synthesizing OpenAI + Anthropic guidance into lectures, projects, and ready-to-copy templates. Core pitch: reliability comes less from a smarter model and more from a closed-loop system, explicit constraints, state management, verification, observability, and control. Newsletter angle: a useful “meta” resource for the current wave of coding-agent practice, especially good if framing the shift from promptcraft to environment/harness design. Retrieval note: extracted cleanly via web_fetch from the lan… https://reading-list.oddship.net/notes/2026-05-12-translation-layer/ https://reading-list.oddship.net/notes/2026-05-12-translation-layer/ Tue, 12 May 2026 00:00:00 +0000 Imported from historical reading log. Blog essay arguing AI compresses the org’s “translation layer” more than any single job title: spec→ticket→PR→release-note work gets cheap, while judgement around why/what/trust systems gets more valuable. Strong claim: middle-management and coordination-heavy roles shrink unless they actively contribute to product definition, architecture, evals, or verification. Newsletter angle: useful framing for how AI changes org shape, not “AI replaces engineers” but “AI eats translation work,” which shifts value toward taste, harnesses, and hands-on decision-maker… https://reading-list.oddship.net/notes/2026-05-07-agent-principal-agent-problem/ https://reading-list.oddship.net/notes/2026-05-07-agent-principal-agent-problem/ Thu, 07 May 2026 00:00:00 +0000 Imported from historical reading log. Read The agent principal-agent problem by David Crawshaw. Core claim: classic review-before-commit code review assumed a human contributor whose effort and understanding could be inferred from the code; agent-mediated contribution breaks that signal and creates a principal-agent problem where reviewers absorb heavy load from low-effort, lightly-validated slop PRs. The useful distinction is not just agents good/bad, but high-trust small teams versus low-trust large organizations: small teams can collapse review and let the human prompter own deployment, wh… https://reading-list.oddship.net/notes/2026-05-07-ai-slop/ https://reading-list.oddship.net/notes/2026-05-07-ai-slop/ Thu, 07 May 2026 00:00:00 +0000 Imported from historical reading log. Extracted main post via api.fxtwitter.com fallback. Mitchell Hashimoto argues that AI slop is useful as an internal experimentation tool: low-quality generated code/UI/plugins can dramatically reduce the cost of parallel exploration and API iteration, especially when regeneration is cheaper than careful hand maintenance. His concrete examples are good: shipping an intentionally rough alpha frontend to focus on core internals, and using overnight agent loops to generate many disposable plugins so the whole ecosystem can be tested before the SDK is stable. … https://reading-list.oddship.net/notes/2026-05-07-claude-mythos-preview/ https://reading-list.oddship.net/notes/2026-05-07-claude-mythos-preview/ Thu, 07 May 2026 00:00:00 +0000 Imported from historical reading log. Extracted Alex Albert's post via api.fxtwitter.com fallback and read the attached chart. Claim: with help from Claude Mythos Preview, the Firefox team fixed more security bugs in April 2026 than in the previous 15 months combined. The screenshot supports the magnitude: Firefox Security Bug Fixes by Month shows a jump from ordinary monthly counts in the ~17–31 range through 2025, then 61 in Feb 2026, 76 in Mar 2026, and a huge spike to 423 in Apr 2026. Caveat worth keeping in mind: the chart is labeled All Sources · All Severities, so this is broader than … https://reading-list.oddship.net/notes/2026-05-07-is-moving-its-github-repo-into-the/ https://reading-list.oddship.net/notes/2026-05-07-is-moving-its-github-repo-into-the/ Thu, 07 May 2026 00:00:00 +0000 Imported from historical reading log. Extracted main post via api.fxtwitter.com fallback. Mario Zechner says pi is moving its GitHub repo into the earendil-works org and will start publishing packages under the @earendil-works npm namespace instead of @mariozechner. Short-term compatibility remains for existing imports, but typed extensions should migrate quickly once the new packages land. Breaking edge: extensions switched to @earendil-works will stop working on older pi versions after today's release. Why it matters: this is an ecosystem/ownership cleanup move, but it deliberately forces e… https://reading-list.oddship.net/notes/2026-05-07-nostalgia-post-in-portuguese-about-the-mid-2000s-pirate-game-install-ritual-utorrent-on-slow-intern/ https://reading-list.oddship.net/notes/2026-05-07-nostalgia-post-in-portuguese-about-the-mid-2000s-pirate-game-install-ritual-utorrent-on-slow-intern/ Thu, 07 May 2026 00:00:00 +0000 Imported from historical reading log. Extracted main post via api.fxtwitter.com fallback. Nostalgia post in Portuguese about the mid-2000s pirate-game install ritual: uTorrent on slow internet, seeding the ISO, Nero burn, Daemon Tools mount, no-CD crack, AVAST warning, mysterious Russian keygen, then finally launching the game. Not really a deep technical claim, but it is a compact cultural artifact of the old PC internet stack: torrents, optical media, disk images, cracks, antivirus false alarms, and the weird literacy that desktop computing once required. Why it matters: useful more as inte… 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-dreaming/ https://reading-list.oddship.net/notes/2026-05-06-dreaming/ Wed, 06 May 2026 00:00:00 +0000 Imported from historical reading log. Claude Managed Agents update centered on three things: dreaming, outcomes, and multiagent orchestration. Dreaming is a research-preview async job that reads an existing memory store plus past session transcripts and emits a cleaned/reorganized memory store with deduped facts, replaced stale entries, and new synthesized insights; original store remains unchanged. Outcomes adds an explicit done target plus rubric-driven grading, turning a session from chat into iterative artifact production with a separate grader context feeding gap reports back to the agen… https://reading-list.oddship.net/notes/2026-05-06-satya-microsoft-framing-firms-need-to-redesign-work-around-agentic-systems-with-ai-taking-more-exec/ https://reading-list.oddship.net/notes/2026-05-06-satya-microsoft-framing-firms-need-to-redesign-work-around-agentic-systems-with-ai-taking-more-exec/ Wed, 06 May 2026 00:00:00 +0000 Imported from historical reading log. Extracted main post via api.fxtwitter.com fallback and read the linked Microsoft Work Trend Index piece Agents, human agency, and the opportunity for organizations. Satya/Microsoft framing: firms need to redesign work around agentic systems, with AI taking more execution while humans shift toward judgment, intent-setting, and owning outcomes. The more interesting claim is organizational, not individual: Microsoft says culture, manager support, and talent practices explain more than 2x the reported AI impact of individual effort alone. Key vocabulary from … 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…