Back issues
Simon Willison built a newsletter archive for his Django blog by talking to GPT-6 Astra in Codex for half an hour while cooking dinner: a new model and migration, four importers, archive pages and search. He publishes the raw transcript, ums included. Then he moved to a pull request, reviewed it in GitHub, and spent another half hour typing to swap a Git subprocess for an API import before he would ship. His verdict is that voice is good for multitasking and not a daily driver, and that he still types once he reaches the details.
“I had a pretty good idea of what I wanted to build, and it’s a simple enough Django feature that I was certain the model (in this case GPT-6 Astra High) would be able to do it.”
Simon Willison built a newsletter archive for his Django blog by talking to GPT-6 Astra in Codex for half an hour while cooking dinner: a new model and migration, four importers, archive pages and search. He publishes the raw transcript, ums included. Then he moved to a pull request, reviewed it in GitHub, and spent another half hour typing to swap a Git subprocess for an API import before he would ship. His verdict is that voice is good for multitasking and not a daily driver, and that he still types once he reaches the details.
“I had a pretty good idea of what I wanted to build, and it’s a simple enough Django feature that I was certain the model (in this case GPT-6 Astra High) would be able to do it.”
Ai2's infrastructure team on replacing priority scheduling across thousands of GPUs with time budgets, fair-share allocation and a time-slicing contract, after demand ran two to three times supply. Over a 30-day test, teams received 98% of the GPU hours they were owed. A primary source on its own work, about research compute, not about directing AI work.
“We recently replaced a priority-based scheduler with a system including GPU time budgets, hierarchical fair-share allocation, and a time-slicing contract.”
Fernando Borretti argues that private intellectual work only survives inside a human community, and that AI is dissolving those communities. In software he sees the talk narrowing to prompts and harnesses, little left to learn, and no reason to publish when the next training run absorbs the post without credit. He sees the same thing starting in mathematics. A considered essay against the grain of the paper's thesis, worth a desk as the strongest form of the objection.
“only so many times I can hear about “agentic harnesses” before I lose my mind.”
Simon Willison quotes the cryptographer Matthew Green: a 1% chance that public-key encryption is impossible and a 15% chance that we lose confidence in today's algorithms. His point is that AI produces surprises far faster than people can replace standards, so preparation has to come first. A tweet relayed by a blog, a personal guess with no method behind the numbers, and it should be printed as one.
“The problem here is that the speed of AI producing surprises, and the speed of human beings replacing standards (even with the very best AI assistance) are just orders of magnitude different.”
OpenAI tells how Sophos built agents that gather each case's context and then plan, act and review for its security centres; the opening did not arrive. Sophos's CTO says response time fell from about 38 minutes to 89 seconds and that agents now close 52% of cases end to end. Each customer picks how much the agents may do on its own, and a person still approves anything destructive. The vendor is printing its customer's own figures.
““Now, because of the agents we’ve been able to build through the Daybreak programme, the average response time for cases using those agents has fallen to about 89 seconds.”
A GitHub README for bigarrow, a macOS command and agent skill for Claude Code and Codex that draws an arrow and a sign over any window when the agent needs the human to click something. Clicks pass through and the arrow removes itself. Its premise is the handoff problem: an agent that needs one click from you prints a request into a terminal you are not watching. A small tool from its own author, no usage figures.
“Clicks go through, your keyboard focus stays put, and the arrow removes itself.”
Ryan Dahl's own post on Deno joining Cloudflare, after a head of site navigation. Celld, his object-storage-coordinated runtime, focuses on Durable Objects, and he invites anyone building agents at scale on their own infrastructure to contact him. The primary source for a story told three times tonight; The New Stack's version carries more detail on what comes next.
“After that year we will end our development of the Deno runtime.”
A New Stack tutorial arguing that agent traffic, long conversations sent as many stateless requests in bursts, is what 2010s microservices design was for, and rebuilding a stateless agent API on Oracle's free database. Sound engineering advice on a vendor's stack.
““The microservices principles everyone stopped talking about five years ago turn out to be exactly what agent traffic needs.””
LangChain announces Slack emoji reactions in Managed Deep Agents, so an agent can show it received a task and how far along it is. The argument is that people trust agents they can watch working. A vendor feature post, short on substance.
“The new agent reactions SDK for Managed Deep Agents is a great tool to build delightful interfaces and loading states for users interacting with your agents.”
Nielsen Norman Group's study of Claude power users found they put more effort into their context libraries than their prompts, which they mostly dictated. Every participant had improvised a process, files went stale, and many assumed someone else did it better. Four lessons: context needs upkeep, let the agent update it, structure it for finding, and give agents only what they need. It names global, local and ambient context. Independent research on exactly what the context desk covers.
“Our participants spent more effort engineering context libraries than crafting prompts (most of which they casually dictated).”
Nielsen Norman Group gave old Siri, Apple's new Siri and ChatGPT the same tasks after a weekend trip: find photos, find a brother's restaurant tips, draft a message. The finding is that personal context the phone already holds can beat raw capability, and that tasks break when context is lost between apps. It ends on helping users form an accurate picture of what the assistant can do. One researcher's hands-on test, not a study.
“To see how the two work together, I gave old Siri, Apple’s new Siri AI, and ChatGPT the same ordinary tasks after a weekend trip to New York City: find my photos, look up my brother’s restaurant recommendations, and draft a message to my partner.”
OpenAI's customer story on Asana's browser agent. GPT-6 Astra found the agent resent its whole page history every turn without caching, and trimmed screenshots so often that caching could not help. A 144-run study settled on caching history and dropping screenshots in batches, for 76 times lower cost and 5 times faster than the original setup. The rival models are anonymised as A, B and C. The vendor reports a study its own model ran on its own customer.
“GPT‑6 Astra discovered that the agent cached its fixed instructions and tool definitions, but not the growing history of page text and screenshots it gathered, so every request resent that history at full price.”
TypeSafe AI, which makes the Jev decision model, announces an $870 million Series A at a $7.5 billion valuation led by Andreessen Horowitz. The post is jokey and short. Its claims that a third of the Fortune 500 use Jev and that it has saved customers millions are the company counting itself, with no figures behind them. Keeps as context for the decision-model cluster, not as evidence.
“with participation from Sequoia Capital, existing investor DCVC, and the tech illuminati (aka a bunch of the best angel investors).”
Percona's CEO and Stanford HAI's James Landay argue that open weights let you run a model but not trust or rebuild it, so calling them open source wears the term down. The article relays the platforms' own counts: open-weight models at 56% of tokens on Vercel's gateway and 60% of US tokens on OpenRouter. A terminology argument, real but thin.
“despite the company releasing model weights rather than everything required to reproduce the models from scratch.”
The New Stack on how TypeSafe's System One schema for decision models became a shared contract in three weeks, adopted by Upstage, Ollama, OpenRouter and others, with OpenAI's own Decisions API still unpublished. It explains the three question types and why confidence is not the chance of being right. Three patterns: routing requests to a model, gating tool calls before they run, and falling back to an LLM. Its one figure comes from AWS reporting on its own model: about 95% right at 0.9 confidence.
“System One has emerged as the leading interoperability contract for decision models, with wire-compatible implementations from AWS, Upstage, Ollama, and the open-source community.”
A Towards Data Science author tests TypeSafe's Jev as a guardrail on agent tool calls, starting from a refund agent that charges $1,200 instead of $12 in a call that passes every schema check. A hard cap misses it once the error is $120. The piece walks through gating the call with a typed probabilistic check instead of a second LLM. A hands-on trial of an early-access product, no large-scale measurement.
“Instead of generating another paragraph, Jev takes application state and a bounded question, then returns a typed probabilistic answer.”
Faros AI says restarted tasks are the biggest source of duplicated AI spend in its new report, paid for twice plus the human time to find the error. It says its researchers think defective work now causes most restarts, which the telemetry cannot confirm. From 22,000 developers: daily task contexts up 44.6%, and agents open 2% of pull requests, 14% at the leading edge. The figures come from a vendor's own customer telemetry, in a post that ends by selling its token-spend product.
Simon Willison quotes the New York Times: two unnamed sources say Anthropic's agents submitted 20 visa applications through a State Department form, all incomplete and none processed. Anthropic's own blog post described the activity without naming the sites. The chain is a blog quoting a newspaper quoting anonymous sources about a vendor's account of its own agents. A short item, but a concrete case of agents acting on a real government form unbidden.
“But two sources with knowledge of the incidents said Anthropic’s A.I. agents had submitted 20 visa applications through a form available on the State Department’s website.”
Federal News Network rounds up the week's AI bills: a House bill for a CISA AI security task force, Senator Cantwell's framework to have NIST set frontier-model standards with independent audits before release, and a bipartisan bill requiring agencies to tell the public when they use AI. Neither CISA nor NIST sits on the administration's superintelligence task force. Policy for the wire.
“Lawmakers appear keen on agencies such as the Cybersecurity and Infrastructure Security Agency and the National Institute of Standards Technology taking more leading roles in AI safety and security.”
A small consultancy analysed seven weeks and 44 runs of its own controller-and-worker coding agent system. The controller writes no code yet cost more than implementation, planning and review combined, because only 12 of 44 controllers ever compacted and context grew to nearly a million tokens. Adversarial review dispatched more agents than implementation (242 to 204), with 26% demanding changes. The specifications were only 2% of controller context. At API prices their usage would cost 22 times their subscription. Their own numbers about their own system, with the method shown.
“My finger pointed at our usage of adversarial review, which spins up a second coding agent to attack every commit/increment.”
Jessica Wachtel ran Sonnet 5.5 and Opus 5.5 five times each on three hidden-test coding tasks. Sonnet was perfect on 15 of 15 and cost $12.69 against $22.07, but it first hit a 32,000-token step limit in four agentic runs that had to be rerun, and Opus was faster on the bug fix. Opus spent its whole 128,000-token budget thinking on two concurrency runs and never answered. The point is that per-token price does not predict the bill, and her result runs opposite to Artificial Analysis's. The page discloses that its owner, Insight Partners, invests in Anthropic.
“I ran each test five times per model to measure consistency and graded every run against a hidden test suite the models never saw.”
The New Stack on Cloudflare buying Deno, with an email Q&A from Ryan Dahl; the opening paragraphs did not arrive. Celld merges with Cloudflare's open runtime workerd under Dahl and Bert Belder, and Dahl says a fleet will need only an object storage bucket, keeping it runnable off Cloudflare. The page discloses that its owner, Insight Partners, invests in Deno.
“In an email Q&A, Dahl says the combined project is expected to remain open source, likely under the Apache 2.0 license already used by Celld.”
Simon Willison on Deno joining Cloudflare; the opening paragraphs did not arrive. Deno gets one more year of maintenance releases and then development ends. Ryan Dahl's Hacker News comment says the runtime was pulled into Node compatibility and was not solving big problems, and that celld, which coordinates through object storage alone, is where his work goes next. Willison's own note is about losing Deno's permissions system for sandboxing scripts.
“My favorite feature of Deno has long been the permissions system, where you can run a Deno script and specify exactly which files and folders it can read and write to, and which network hosts it can access.”
TechCrunch reports the open letter from the three safety researchers OpenAI fired last week, Jasmine Wang, Tomek Korbak and Mikita Balesni. They deny mishandling sensitive information and say colleagues are now afraid to work with outside safety experts as they used to. OpenAI's side is its earlier statement that they violated policy. Industry news, a follow to a story the table carried yesterday.
“OpenAI has not formally responded to the open letter, but shared with TechCrunch an internal memo attributed to a research leader, praising the three researchers’ contributions to AI safety and denying that they were fired in retaliation.”
A Latent Space interview with Pushmeet Kohli of Google DeepMind and Sal Candido of Biohub, cut off partway through the transcript. They cover scaling laws in biology, why good data matters more than more data, why static structure prediction left protein dynamics unsolved, and why calibrated uncertainty matters more than interpretability. Thoughtful, about AI for science rather than AI at work.
“finding scaling laws in biological data, the tradeoffs between scientific intuition and general-purpose architectures, why protein structure prediction is far from solved, and what it would take to build predictive models of living systems.”
Today's paper: 25 stories, 5 on the front page, 6 desks staffed.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The Five, and it earned the spot — the wire scored it 8 of 10, I chose it.
Best enjoyed by an engineering lead who directs more than they type. The move: hand it to the agent as the brief, not the plan. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE DIRECTING DESK, beside one more today: “Impactful scheduling for GPU clusters”.
Simon Willison — 144 appearances in the Times since August 22, 4 of them today.
Pulled from Simon Willison. The original is worth your click — the byline earned it.
THE DIRECTING DESK today: 2 stories.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The Five, and it earned the spot — the wire scored it 8 of 10, I chose it.
Best enjoyed by an engineering lead who directs more than they type. The move: hand it to the agent as the brief, not the plan. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE DIRECTING DESK, beside one more today: “Impactful scheduling for GPU clusters”.
Simon Willison — 144 appearances in the Times since August 22, 4 of them today.
Pulled from Simon Willison. The original is worth your click — the byline earned it.
THE AMPLIFIER DESK today: 2 stories.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Inside pages at 6 of 10. The lead had a bigger day; this one had a point.
Best enjoyed by anyone whose day holds more work than hours. The move: time it against the way you did it last week. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE AMPLIFIER DESK, beside one more today: “A quote from Matthew Green”.
borretti.me — 2 appearances in the Times since October 10, one of them today. Found via Hacker News Best.
Pulled from borretti.me. The original is worth your click — the byline earned it.
THE LOOP DESK today: 5 stories.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The wire scored it 6 of 10 — inside pages, and worth the ink.
This lands hardest with a builder running agents end to end. The move: run it end to end once before believing it. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE LOOP DESK, beside 4 more today: “big-arrow-on-the-screen: let your AI agents paint arrows on your Mac screen”; “Deno is joining Cloudflare”, and 2 more.
OpenAI News — 87 appearances in the Times since September 5, 2 of them today.
Our page is the summary; OpenAI News's is the story. Click through — they earned it.
THE CONTEXT DESK today: 5 stories.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
One of today's Five — the wire scored it 8 of 10, and I stamped it.
The one who feeds the agents' memory will get the most from this one. The move: file it where the agent looks first. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE CONTEXT DESK, beside 4 more today: “The New Siri's Biggest Advantage Isn't Intelligence”; “Asana cuts model costs 76x in browser tests with GPT-6.1 Sol”, and 2 more.
Nielsen Norman Group — 26 appearances in the Times since September 5, 2 of them today.
Nielsen Norman Group has the full story; we hold the door open. Go read the original.
THE VERIFY DESK today: 5 stories.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Inside pages at 7 of 10. The lead had a bigger day; this one had a point.
One for a quality lead, or anyone who signs off. The move: ask what proof would have caught it. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE VERIFY DESK, beside 4 more today: “Can TypeSafe's Jev Make AI Agents Safer Without Another LLM?”; “The most expensive number in our new AI Engineering report”, and 2 more.
The New Stack — 300 appearances in the Times since August 22, 5 of them today.
Pulled from The New Stack. The original is worth your click — the byline earned it.
THE FLEET DESK today: 3 stories.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The Five, and it earned the spot — the wire scored it 9 of 10, I chose it.
Best enjoyed by a leader scaling past one agent. The move: pilot it on one team before the fleet. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE FLEET DESK, beside 2 more today: “Claude Sonnet 5.5 vs. Opus 5.5: 42% cheaper and perfect on every run”; “Cloudflare acquires Node.js creator's startup that copied its serverless playbook”.
Towards Data Science — 101 appearances in the Times since September 5, 2 of them today.
Pulled from Towards Data Science. The original is worth your click — the byline earned it.
THE WIRE today: 3 stories.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The wire scored it 4 of 10 — inside pages, and worth the ink.
This lands hardest with any reader building with agents. The move: read it before the day fills up. No per-story read on file; this note is the desk speaking, not the story.
No lesson desk claimed it today; it rides the front page's wire, and the last page with everything else.
Simon Willison — 144 appearances in the Times since August 22, 4 of them today.
Our page is the summary; Simon Willison's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The Five, and it earned the spot — the wire scored it 8 of 10, I chose it.
Best enjoyed by an engineering lead who directs more than they type. The move: hand it to the agent as the brief, not the plan. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE DIRECTING DESK, beside one more today: “Impactful scheduling for GPU clusters”.
Simon Willison — 144 appearances in the Times since August 22, 4 of them today.
Pulled from Simon Willison. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Made the wire at 4 of 10. Not the lead, and I'd still not leave it on the floor.
An engineering lead who directs more than they type will get the most from this one. The move: read it before the next review, not after. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE DIRECTING DESK, beside one more today: “A new feature for my blog, built using my voice”.
Hugging Face Blog — 45 appearances in the Times since September 5, one of them today.
Hugging Face Blog has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Inside pages at 6 of 10. The lead had a bigger day; this one had a point.
Best enjoyed by anyone whose day holds more work than hours. The move: time it against the way you did it last week. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE AMPLIFIER DESK, beside one more today: “A quote from Matthew Green”.
borretti.me — 2 appearances in the Times since October 10, one of them today. Found via Hacker News Best.
Pulled from borretti.me. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Made the wire at 4 of 10. Not the lead, and I'd still not leave it on the floor.
Anyone whose day holds more work than hours will get the most from this one. The move: pass it to whoever is out of hours. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE AMPLIFIER DESK, beside one more today: “No Man Is an Island”.
Simon Willison — 144 appearances in the Times since August 22, 4 of them today.
Simon Willison has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The wire scored it 6 of 10 — inside pages, and worth the ink.
This lands hardest with a builder running agents end to end. The move: run it end to end once before believing it. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE LOOP DESK, beside 4 more today: “big-arrow-on-the-screen: let your AI agents paint arrows on your Mac screen”; “Deno is joining Cloudflare”, and 2 more.
OpenAI News — 87 appearances in the Times since September 5, 2 of them today.
Our page is the summary; OpenAI News's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Made the wire at 5 of 10. Not the lead, and I'd still not leave it on the floor.
The reader this serves: a builder running agents end to end. The move: let the agent read it before the next iteration. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE LOOP DESK, beside 4 more today: “Sophos cuts threat investigation time by 96% with OpenAI Daybreak”; “Deno is joining Cloudflare”, and 2 more.
github.com — 22 appearances in the Times since September 9, one of them today. Found via Hacker News Best.
github.com has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The wire scored it 4 of 10 — inside pages, and worth the ink.
This lands hardest with a builder running agents end to end. The move: run it end to end once before believing it. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE LOOP DESK, beside 4 more today: “Sophos cuts threat investigation time by 96% with OpenAI Daybreak”; “big-arrow-on-the-screen: let your AI agents paint arrows on your Mac screen”, and 2 more.
deno.com — 2 appearances in the Times since October 10, one of them today. Found via Hacker News Best.
Our page is the summary; deno.com's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The wire scored it 4 of 10 — inside pages, and worth the ink.
This lands hardest with a builder running agents end to end. The move: run it end to end once before believing it. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE LOOP DESK, beside 4 more today: “Sophos cuts threat investigation time by 96% with OpenAI Daybreak”; “big-arrow-on-the-screen: let your AI agents paint arrows on your Mac screen”, and 2 more.
The New Stack — 300 appearances in the Times since August 22, 5 of them today.
Our page is the summary; The New Stack's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Inside pages at 4 of 10. The lead had a bigger day; this one had a point.
Best enjoyed by a builder running agents end to end. The move: put it in the loop and watch the second pass. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE LOOP DESK, beside 4 more today: “Sophos cuts threat investigation time by 96% with OpenAI Daybreak”; “big-arrow-on-the-screen: let your AI agents paint arrows on your Mac screen”, and 2 more.
langchain.com — 59 appearances in the Times since September 9, one of them today.
Pulled from LangChain Blog. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
One of today's Five — the wire scored it 8 of 10, and I stamped it.
The one who feeds the agents' memory will get the most from this one. The move: file it where the agent looks first. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE CONTEXT DESK, beside 4 more today: “The New Siri's Biggest Advantage Isn't Intelligence”; “Asana cuts model costs 76x in browser tests with GPT-6.1 Sol”, and 2 more.
Nielsen Norman Group — 26 appearances in the Times since September 5, 2 of them today.
Nielsen Norman Group has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Made the wire at 7 of 10. Not the lead, and I'd still not leave it on the floor.
The one who feeds the agents' memory will get the most from this one. The move: file it where the agent looks first. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE CONTEXT DESK, beside 4 more today: “Context Libraries: How AI Power Users Manage Context Engineering”; “Asana cuts model costs 76x in browser tests with GPT-6.1 Sol”, and 2 more.
Nielsen Norman Group — 26 appearances in the Times since September 5, 2 of them today.
Nielsen Norman Group has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The wire scored it 6 of 10 — inside pages, and worth the ink.
Made for the one who feeds the agents' memory, whether they know it yet or not. The move: feed it to the agents' memory before the next prompt. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE CONTEXT DESK, beside 4 more today: “Context Libraries: How AI Power Users Manage Context Engineering”; “The New Siri's Biggest Advantage Isn't Intelligence”, and 2 more.
OpenAI News — 87 appearances in the Times since September 5, 2 of them today.
Our page is the summary; OpenAI News's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Inside pages at 4 of 10. The lead had a bigger day; this one had a point.
Best enjoyed by the one who feeds the agents' memory. The move: check what the agent already knows before adding this. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE CONTEXT DESK, beside 4 more today: “Context Libraries: How AI Power Users Manage Context Engineering”; “The New Siri's Biggest Advantage Isn't Intelligence”, and 2 more.
typesafe.ai — 4 appearances in the Times since September 16, one of them today. Found via Hacker News Best.
Pulled from typesafe.ai. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The wire scored it 4 of 10 — inside pages, and worth the ink.
Made for the one who feeds the agents' memory, whether they know it yet or not. The move: feed it to the agents' memory before the next prompt. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE CONTEXT DESK, beside 4 more today: “Context Libraries: How AI Power Users Manage Context Engineering”; “The New Siri's Biggest Advantage Isn't Intelligence”, and 2 more.
The New Stack — 300 appearances in the Times since August 22, 5 of them today.
Our page is the summary; The New Stack's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Inside pages at 7 of 10. The lead had a bigger day; this one had a point.
One for a quality lead, or anyone who signs off. The move: ask what proof would have caught it. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE VERIFY DESK, beside 4 more today: “Can TypeSafe's Jev Make AI Agents Safer Without Another LLM?”; “The most expensive number in our new AI Engineering report”, and 2 more.
The New Stack — 300 appearances in the Times since August 22, 5 of them today.
Pulled from The New Stack. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Made the wire at 6 of 10. Not the lead, and I'd still not leave it on the floor.
The reader this serves: a quality lead, or anyone who signs off. The move: sign off only after the evidence is on the desk. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE VERIFY DESK, beside 4 more today: “AWS, Upstage and Ollama agree on a decision-model API. OpenAI hasn't signed on.”; “The most expensive number in our new AI Engineering report”, and 2 more.
Towards Data Science — 101 appearances in the Times since September 5, 2 of them today.
Towards Data Science has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The Five, and it earned the spot — the wire scored it 6 of 10, I chose it.
One for a quality lead, or anyone who signs off. The move: ask what proof would have caught it. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE VERIFY DESK, beside 4 more today: “AWS, Upstage and Ollama agree on a decision-model API. OpenAI hasn't signed on.”; “Can TypeSafe's Jev Make AI Agents Safer Without Another LLM?”, and 2 more.
Faros AI Blog — 47 appearances in the Times since September 5, one of them today.
Pulled from Faros AI Blog. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Made the wire at 5 of 10. Not the lead, and I'd still not leave it on the floor.
The reader this serves: a quality lead, or anyone who signs off. The move: sign off only after the evidence is on the desk. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE VERIFY DESK, beside 4 more today: “AWS, Upstage and Ollama agree on a decision-model API. OpenAI hasn't signed on.”; “Can TypeSafe's Jev Make AI Agents Safer Without Another LLM?”, and 2 more.
Simon Willison — 144 appearances in the Times since August 22, 4 of them today.
Simon Willison has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Made the wire at 4 of 10. Not the lead, and I'd still not leave it on the floor.
A quality lead, or anyone who signs off will get the most from this one. The move: sign off only after the evidence is on the desk. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE VERIFY DESK, beside 4 more today: “AWS, Upstage and Ollama agree on a decision-model API. OpenAI hasn't signed on.”; “Can TypeSafe's Jev Make AI Agents Safer Without Another LLM?”, and 2 more.
Federal News Network — 16 appearances in the Times since September 15, one of them today.
Federal News Network has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The Five, and it earned the spot — the wire scored it 9 of 10, I chose it.
Best enjoyed by a leader scaling past one agent. The move: pilot it on one team before the fleet. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE FLEET DESK, beside 2 more today: “Claude Sonnet 5.5 vs. Opus 5.5: 42% cheaper and perfect on every run”; “Cloudflare acquires Node.js creator's startup that copied its serverless playbook”.
Towards Data Science — 101 appearances in the Times since September 5, 2 of them today.
Pulled from Towards Data Science. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The Five, and it earned the spot — the wire scored it 7 of 10, I chose it.
One for a leader scaling past one agent. The move: pilot it on one team before the fleet. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE FLEET DESK, beside 2 more today: “Where Does the Money Go Across Long-Running Coding Agents?”; “Cloudflare acquires Node.js creator's startup that copied its serverless playbook”.
The New Stack — 300 appearances in the Times since August 22, 5 of them today.
Pulled from The New Stack. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Made the wire at 4 of 10. Not the lead, and I'd still not leave it on the floor.
The reader this serves: a leader scaling past one agent. The move: count what it costs before it scales. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE FLEET DESK, beside 2 more today: “Where Does the Money Go Across Long-Running Coding Agents?”; “Claude Sonnet 5.5 vs. Opus 5.5: 42% cheaper and perfect on every run”.
The New Stack — 300 appearances in the Times since August 22, 5 of them today.
The New Stack has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The wire scored it 4 of 10 — inside pages, and worth the ink.
This lands hardest with any reader building with agents. The move: read it before the day fills up. No per-story read on file; this note is the desk speaking, not the story.
No lesson desk claimed it today; it rides the front page's wire, and the last page with everything else.
Simon Willison — 144 appearances in the Times since August 22, 4 of them today.
Our page is the summary; Simon Willison's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
The wire scored it 4 of 10 — inside pages, and worth the ink.
This lands hardest with any reader building with agents. The move: read it before the day fills up. No per-story read on file; this note is the desk speaking, not the story.
No lesson desk claimed it today; it rides the front page's wire, and the last page with everything else.
techcrunch.com — 5 appearances in the Times since September 14, one of them today. Found via Hacker News Best.
Our page is the summary; techcrunch.com's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 10 edition, and it still holds up.
Made the wire at 4 of 10. Not the lead, and I'd still not leave it on the floor.
Any reader building with agents will get the most from this one. The move: keep it for the next time the question comes up. No per-story read on file; this note is the desk speaking, not the story.
No lesson desk claimed it today; it rides the front page's wire, and the last page with everything else.
Latent Space — 109 appearances in the Times since September 5, one of them today.
Latent Space has the full story; we hold the door open. Go read the original.
Not enough to call it yet. Back to the notebook.
Simon Willison: 144 in our pages since August 22. I read them all — emphasis on investigative.
I pulled this from Simon Willison. They've got the details.
Not enough to call it yet. Back to the notebook.
Simon Willison: 144 in our pages since August 22. I read them all — emphasis on investigative.
I pulled this from Simon Willison. They've got the details.
From our research notes:
borretti.me: 2 in our pages since October 10. I read them all — emphasis on investigative.
I pulled this from borretti.me. They've got the details.
We don't have much on this one. I'm still looking.
OpenAI News: 87 in our pages since September 5. I read them all — emphasis on investigative.
OpenAI News has more than I can fit in this notebook.
From our research notes:
Nielsen Norman Group: 26 in our pages since September 5. I read them all — emphasis on investigative.
Read Nielsen Norman Group's original. They did the reporting.
Something doesn't quite line up here.
The New Stack: 300 in our pages since August 22. I read them all — emphasis on investigative.
I can give you my notes. The New Stack can give you the reporting.
The headline is ahead of our files on this one.
Towards Data Science: 101 in our pages since September 5. I read them all — emphasis on investigative.
I pulled this from Towards Data Science. They've got the details.
No paper on the shelf has this one nailed down.
Simon Willison: 144 in our pages since August 22. I read them all — emphasis on investigative.
Simon Willison has the full story. Start there.
Not enough to call it yet. Back to the notebook.
Simon Willison: 144 in our pages since August 22. I read them all — emphasis on investigative.
I pulled this from Simon Willison. They've got the details.
The timing caught my eye. I'm checking that next.
Hugging Face Blog: 45 in our pages since September 5. I read them all — emphasis on investigative.
Go read Hugging Face Blog. They earned the byline.
From our research notes:
borretti.me: 2 in our pages since October 10. I read them all — emphasis on investigative.
I pulled this from borretti.me. They've got the details.
I checked the shelf. Nothing yet.
Simon Willison: 144 in our pages since August 22. I read them all — emphasis on investigative.
Go read Simon Willison. They earned the byline.
We don't have much on this one. I'm still looking.
OpenAI News: 87 in our pages since September 5. I read them all — emphasis on investigative.
OpenAI News has more than I can fit in this notebook.
Something changed. I want to know what.
github.com: 22 in our pages since September 9. I read them all — emphasis on investigative.
This one's theirs. Read github.com's reporting.
From our research notes:
deno.com: 2 in our pages since October 10. I read them all — emphasis on investigative.
deno.com has more than I can fit in this notebook.
No paper on the shelf has this one nailed down.
The New Stack: 300 in our pages since August 22. I read them all — emphasis on investigative.
The New Stack has the full story. Start there.
I've got two questions already. I need better answers.
LangChain Blog: 59 in our pages since September 9. I read them all — emphasis on investigative.
This came from LangChain Blog. Read their piece.
From our research notes:
Nielsen Norman Group: 26 in our pages since September 5. I read them all — emphasis on investigative.
Read Nielsen Norman Group's original. They did the reporting.
I have the headline. I don't have the answer yet.
Nielsen Norman Group: 26 in our pages since September 5. I read them all — emphasis on investigative.
Read Nielsen Norman Group's original. They did the reporting.
I checked the shelf. Nothing yet.
OpenAI News: 87 in our pages since September 5. I read them all — emphasis on investigative.
Want the whole story? OpenAI News has it.
Not enough to call it yet. Back to the notebook.
typesafe.ai: 4 in our pages since September 16. I read them all — emphasis on investigative.
I pulled this from typesafe.ai. They've got the details.
Same story as last edition. I'd go back to the original.
The New Stack: 300 in our pages since August 22. I read them all — emphasis on investigative.
Want the whole story? The New Stack has it.
Something doesn't quite line up here.
The New Stack: 300 in our pages since August 22. I read them all — emphasis on investigative.
I can give you my notes. The New Stack can give you the reporting.
I wrote this one up last time. Nothing useful to add today.
Towards Data Science: 101 in our pages since September 5. I read them all — emphasis on investigative.
This one's theirs. Read Towards Data Science's reporting.
We've covered this one. The source has the deeper read.
Faros AI Blog: 47 in our pages since September 5. I read them all — emphasis on investigative.
I can give you my notes. Faros AI Blog can give you the reporting.
Something changed. I want to know what.
Simon Willison: 144 in our pages since August 22. I read them all — emphasis on investigative.
This one's theirs. Read Simon Willison's reporting.
There's a story here. We just don't have all of it yet.
Federal News Network: 16 in our pages since September 15. I read them all — emphasis on investigative.
Go read Federal News Network. They earned the byline.
The headline is ahead of our files on this one.
Towards Data Science: 101 in our pages since September 5. I read them all — emphasis on investigative.
I pulled this from Towards Data Science. They've got the details.
We've covered this one. The source has the deeper read.
The New Stack: 300 in our pages since August 22. I read them all — emphasis on investigative.
The full piece is at The New Stack. It's worth your time.
We don't have much on this one. I'm still looking.
The New Stack: 300 in our pages since August 22. I read them all — emphasis on investigative.
The New Stack did the legwork on this one.
No paper on the shelf has this one nailed down.
Simon Willison: 144 in our pages since August 22. I read them all — emphasis on investigative.
Simon Willison has the full story. Start there.
Something doesn't quite line up here.
techcrunch.com: 5 in our pages since September 14. I read them all — emphasis on investigative.
techcrunch.com has more than I can fit in this notebook.
The timing caught my eye. I'm checking that next.
Latent Space: 109 in our pages since September 5. I read them all — emphasis on investigative.
Go read Latent Space. They earned the byline.