Back issues
LangChain raised its coding agent from 52.8 to 66.5 on Terminal Bench 2.0 by changing only the harness, with the model held fixed at gpt-5.2-codex. It read traces to find failure modes and tuned three knobs: system prompt, tools and middleware. A vendor measuring itself on a public benchmark with the method shown. It names gpt-5.2-codex, so it reads as an archive post re-sent by the feed.
A Willison beat: he asked Opus 5.5 for a three.js spot-the-difference game and told it to make sure it was fun. He says that instruction works uncomfortably well; the model ran for three hours and seven minutes and came back with music, animation and achievements. A short first-person note with no measurement, but a plain example of stating the goal and letting the agent run.
About two dozen House Democrats ask GAO to audit the FAA's new AI-assisted air traffic control tools under the modernisation push, citing questions about reliability and safety. A letter asking for an independent check on AI in a safety-critical system; real reporting, no findings yet.
Google Cloud on daily token quotas for BigQuery's LLM functions, per project and per user, so one SQL query scoring millions of rows cannot run up an unbounded bill. Vendor documentation, but it is the hard budget cap Willison asked for earlier this month, shipped as a default.
Latent Space's AINews roundup from AI Engineer New York, careful about whose numbers are whose. Google's 35% more submissions and a third fewer rollbacks are labelled internal claims, not controlled measurements. Artificial Analysis puts Argon level with GPT-6 Astra at about 60% of its cost, with prices doubling after November 30. Theo reports a false-claim rate above 50% on his codebase-analysis test despite a perfect score elsewhere. Mollick finds weaker multi-agent execution and overconfidence. The body stops at the cap.
METR on how an assessor could tell whether automating AI research is speeding up capability progress: benchmarks, intermediate measures like pretraining loss, researcher surveys, and how much of any speed-up comes from algorithms rather than compute. A primary-source proposal, further from day-to-day practice than its three siblings.
LangChain raised its coding agent from 52.8 to 66.5 on Terminal Bench 2.0 by changing only the harness, with the model held fixed at gpt-5.2-codex. It read traces to find failure modes and tuned three knobs: system prompt, tools and middleware. A vendor measuring itself on a public benchmark with the method shown. It names gpt-5.2-codex, so it reads as an archive post re-sent by the feed.
Claire Vo ran Meta's Muse Wonder models through her own How I AI bench of about twenty tasks, scored by an LLM judge and her notes. The page is the episode's outline: strong on three agentic tasks, poor at design, slow on agent work. The results themselves are in the video, not the body.
GitHub's availability report for August and September: 120 of 160 services and 13 of 51 primary databases moved to Azure's Central US, a capacity pause on September 11, and a missed migration target. The September 13 incident is worth a read. A cleanup job's safeguard watched replica lag but not load on the primary, there was no fast timeout, and a retry loop kept the database saturated. A primary source on its own operations; it never says how much of the load is agents.
Pragmatic Engineer with an Uber Distinguished Engineer on its agent tooling. Every model call goes through one gateway with security checks and per-user spend tracking, and token costs fell about half in recent months. CI jobs more than doubled in a year, forcing a code red and moving lint and build checks into the agent's own loop. Code review is the admitted bottleneck, with AI review and risk scores before a human reads. PRs are now tagged by type because PR counts stopped meaning anything. His summary of the job: build the right context so agents produce reliable output, and verify it. A company describing itself, with numbers but no method shown.
LangChain on context engineering in its Deep Agents framework, worked through a weekly competitor-report agent that by step four is carrying five raw exports, dozens of search results and two stale drafts into every call. Retrieval only solves half of it; the harness also has to offload, summarise and drop what is already in the window, done as middleware hooked around model and tool calls. A clear worked example from a vendor describing its own framework.
Subquadratic says its Snap decision model reads 100,000 tokens and picks from fixed answers in about a second, so an agent can hand routine choices such as the next tool, the relevant file or whether a test passed to a small model. The figures are the company's own, and the CEO says the design follows a rival's. On the cost-routing idea the paper has been tracking; no independent measurement.
LangChain adds ScaleDown's small constrained-decoding models for the routine context decisions a long-running agent makes, summarising old history, compressing tool results and extracting spans, without a frontier model. The 200x faster and 467x cheaper figures are ScaleDown's own, relayed by its integration partner. A partner launch on the cheap-model-for-easy-decisions idea.
Chainguard launches Harness, a Claude Mythos-powered bug hunter it built for its own Project Glasswing work. A finding only counts if the exploit reproduces in a fresh isolated environment, and a fix only moves if the exploit fails against it; an engineer approves every report. Its first public finding is a vLLM flaw where one client could tamper with another's output silently. The same launch adds per-agent microVMs, credentials kept outside the agent and CI checks the agent runs before opening a PR. The 66% headline is a Chainguard-commissioned survey, and the product claims are the vendor's.
An OpenAI customer story with Block's CISO, more concrete than most. Every investigation stores its raw query evidence, so a detection change can be replayed against seven months of telemetry in under an hour. The harness routes tasks to different models, lets agents check each other's work and hands off to humans. Every finding goes propose patch, confirm exploitable, verify fixed, with a human deploying. The figures (1,100 analyst-hours a quarter, 50% more triage, 18 minutes from finding to proposed patch in a red-team exercise) are Block's own on OpenAI's page.
Latent Space's report from AI Engineer New York on how finance runs agents. Two Sigma warns of approval fatigue and uses classifiers so people see only the calls that matter. Anthropic's applied team describes identity, sandboxed approval and an agent that proposes order amendments but never makes them. Bloomberg's architecture head describes an agent that took a marketplace review as an instruction to stock a product, and wants deterministic controls outside the model: check every call, log it, keep a kill switch. Conference keynotes reported secondhand, with the speakers named.
METR's proposal for giving the public evidence on whether labs could lose control of their AI systems: the core questions about internal development, and the access a lab would grant an independent assessor to answer them. It says outright that it offers no assurance, since the science is immature and there is no standard to audit against, only new evidence for others to judge. The body starts partway in and stops at the cap.
LangChain answers the common questions about agent evals: start with one behaviour and a small dataset built from real traces, run offline evals before shipping and online ones after, and turn each new failure into a test case. Use code checks where possible, LLM judges where not, and human review over both; for multi-step agents, grade the whole trajectory and the final state, not only the answer. Sound practice, written by the vendor of the eval platform it recommends; the body stops at the cap.
OpenCode's own 2.0 release notes. The agent now runs as one background service that the terminal, web and desktop clients share, with a rebuilt API, background subagents (one can be a different model, for example a reviewer), async shell commands and a code mode that turns MCP servers into functions the agent calls from a script. A vendor describing its own product, with no numbers beyond the claim of millions of daily users.
LangChain releases Open SWE, an open-source framework for internal coding agents built from patterns at Stripe, Ramp and Coinbase: isolated cloud sandboxes with full permissions inside hard limits, a curated toolset (Stripe's roughly 500 tools are maintained, not accumulated), subagents and Slack-first use. A useful summary of others' designs packaged as the vendor's product, and it reads as an older post the feed re-sent.
OpenAI Economic Research on agentic use in the EU, from its own customer data. Codex and ChatGPT Work produce more than half of output tokens in its EU enterprise sample, weekly users of those tools at small and medium firms grew more than eightfold since May, and the agentic share of tokens runs from 22% in Slovenia to 85% in Lithuania. A vendor counting its own customers, with the sample limits stated.
A former Meta source-control engineer says Anthropic's post on rebuilding its source control credited Sapling only for lessons learned and never named Mononoke, the Meta system whose design it follows, and calls that close to plagiarism. The body the net kept is the post's footnote, not the post. The point worth keeping: models asked about prior art name Mononoke at once, so crediting earlier work is now easy and has no excuse. The relay of Anthropic's post was scored last night; the primary was refused by our guard.
Willison links an indie developer's essay on AI-driven decompilation of games. The quoted part covers both sides: anyone can now decompile and reuse paid software, and old games become far easier to port and preserve. A link post with his short framing, so the argument is the original author's.
JPMorgan, Morgan Stanley and two Japanese banks are looking to shift the risk on AI data-centre loans they underwrote months ago, as they near internal limits on AI exposure. A $38 billion Oracle-leased project has gone unsold for six months, and Oracle's credit default swaps are at levels last seen in 2009. A startup news site relaying the Financial Times and Bloomberg; the reporting is theirs.
An OpenAI customer story: SentinelOne's chief AI officer says OpenAI's Daybreak produces human-level malware reverse-engineering in about eight hours, against four to six weeks for a top researcher. He also says the one thing that cannot be automated is accountability. The figure is the customer's estimate on the vendor's page, with no method.
Stack Overflow interviews Andela's head of research on nearly 50,000 job postings read against its skills taxonomy to find roles emerging before they have titles. An answer to the AI-and-jobs talk with data behind it, though the method is the firm's own and only sketched; the body stops at the cap.
Today's paper: 25 stories, 5 on the front page, 6 desks staffed.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
Today's Five, front page: the wire scored it 6 of 10; the stamp is mine.
Made for a builder running agents end to end, whether they know it yet or not. 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 2 more today: “Meta’s new ‘Muse Wonder’ model: good for AI agents, bad for design”; “GitHub availability report: September 2026”.
langchain.com — 59 appearances in the Times since September 9, 5 of them today.
Our page is the summary; LangChain Blog's is the story. Click through — they earned it.
THE DIRECTING DESK today: 3 stories.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
Inside pages at 4 of 10. The lead had a bigger day; this one had a point.
One for 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 2 more today: “House Democrats seek watchdog review of AI-powered air traffic control systems”; “Using BigQuery’s generative AI cost management features | Google Cloud Blog”.
Simon Willison — 144 appearances in the Times since August 22, 2 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 15 edition, and it still holds up.
Inside pages at 6 of 10. The lead had a bigger day; this one had a point.
One for 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: “How an assessor could investigate evidence of acceleration from the automation of AI R&D”.
Latent Space — 109 appearances in the Times since September 5, 2 of them today.
Pulled from Latent Space. The original is worth your click — the byline earned it.
THE LOOP DESK today: 3 stories.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
Today's Five, front page: the wire scored it 6 of 10; the stamp is mine.
Made for a builder running agents end to end, whether they know it yet or not. 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 2 more today: “Meta’s new ‘Muse Wonder’ model: good for AI agents, bad for design”; “GitHub availability report: September 2026”.
langchain.com — 59 appearances in the Times since September 9, 5 of them today.
Our page is the summary; LangChain Blog's is the story. Click through — they earned it.
THE CONTEXT DESK today: 4 stories.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
The Five, and it earned the spot — the wire scored it 8 of 10, I chose it.
One for 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 3 more today: “Context Engineering with Deep Agents: How To Remember Without Keeping Everything In Context”; ““Take work away from frontier LLMs”: Subquadratic’s decision model claims 12x Jev’s context window”, and 1 more.
Pragmatic Engineer — 36 appearances in the Times since September 5, one of them today.
Pulled from Pragmatic Engineer. The original is worth your click — the byline earned it.
THE VERIFY DESK today: 5 stories.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
Inside pages at 7 of 10. The lead had a bigger day; this one had a point.
Best enjoyed by 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: “How Block is building AI-powered security with OpenAI technology”; “Crossing the Threshold: How OpenAI and Anthropic Approach Autonomy and Safety in Financial Agents”, and 2 more.
The New Stack — 300 appearances in the Times since August 22, 2 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 15 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 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: “Open SWE: An Open-Source Framework for Internal Coding Agents”; “The EU’s next chapter in AI is agentic”.
anoma.ly — 2 appearances in the Times since October 15, one of them today. Found via Hacker News Best.
anoma.ly has the full story; we hold the door open. Go read the original.
THE WIRE today: 5 stories.
The presses were quiet this morning — this is the October 15 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: any reader building with agents. 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.
sunshowers.io — 4 appearances in the Times since October 15, one of them today. Found via Hacker News Best.
sunshowers.io has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
Inside pages at 4 of 10. The lead had a bigger day; this one had a point.
One for 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 2 more today: “House Democrats seek watchdog review of AI-powered air traffic control systems”; “Using BigQuery’s generative AI cost management features | Google Cloud Blog”.
Simon Willison — 144 appearances in the Times since August 22, 2 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 15 edition, and it still holds up.
Inside pages at 4 of 10. The lead had a bigger day; this one had a point.
One for 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 2 more today: “Tool: Tree Appreciator”; “Using BigQuery’s generative AI cost management features | Google Cloud Blog”.
Nextgov AI — 48 appearances in the Times since September 5, one of them today.
Pulled from Nextgov AI. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
The wire scored it 4 of 10 — inside pages, and worth the ink.
This lands hardest with an engineering lead who directs more than they type. The move: rewrite Monday's first prompt with it. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE DIRECTING DESK, beside 2 more today: “Tool: Tree Appreciator”; “House Democrats seek watchdog review of AI-powered air traffic control systems”.
Google Cloud AI Blog — 71 appearances in the Times since September 7, one of them today.
Our page is the summary; Google Cloud AI Blog's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
Inside pages at 6 of 10. The lead had a bigger day; this one had a point.
One for 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: “How an assessor could investigate evidence of acceleration from the automation of AI R&D”.
Latent Space — 109 appearances in the Times since September 5, 2 of them today.
Pulled from Latent Space. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 15 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.
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: “[AINews] AI Engineer NYC: Gemini 4 Argon, Vals AI Benchmarks”.
METR — 22 appearances in the Times since September 6, 2 of them today.
METR has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
Today's Five, front page: the wire scored it 6 of 10; the stamp is mine.
Made for a builder running agents end to end, whether they know it yet or not. 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 2 more today: “Meta’s new ‘Muse Wonder’ model: good for AI agents, bad for design”; “GitHub availability report: September 2026”.
langchain.com — 59 appearances in the Times since September 9, 5 of them today.
Our page is the summary; LangChain Blog's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 15 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 2 more today: “Improving Deep Agents with harness engineering”; “GitHub availability report: September 2026”.
Lenny's Newsletter — 54 appearances in the Times since August 24, one of them today.
Pulled from Lenny's Newsletter. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 15 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 builder running agents end to end will get the most from this one. 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 2 more today: “Improving Deep Agents with harness engineering”; “Meta’s new ‘Muse Wonder’ model: good for AI agents, bad for design”.
GitHub Blog AI — 40 appearances in the Times since September 5, one of them today.
GitHub Blog AI has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
The Five, and it earned the spot — the wire scored it 8 of 10, I chose it.
One for 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 3 more today: “Context Engineering with Deep Agents: How To Remember Without Keeping Everything In Context”; ““Take work away from frontier LLMs”: Subquadratic’s decision model claims 12x Jev’s context window”, and 1 more.
Pragmatic Engineer — 36 appearances in the Times since September 5, one of them today.
Pulled from Pragmatic Engineer. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 15 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: the one who feeds the agents' memory. 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 3 more today: “Uber’s AI infra, from the inside”; ““Take work away from frontier LLMs”: Subquadratic’s decision model claims 12x Jev’s context window”, and 1 more.
langchain.com — 59 appearances in the Times since September 9, 5 of them today.
LangChain Blog has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 15 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 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 3 more today: “Uber’s AI infra, from the inside”; “Context Engineering with Deep Agents: How To Remember Without Keeping Everything In Context”, and 1 more.
The New Stack — 300 appearances in the Times since August 22, 2 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 15 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 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 3 more today: “Uber’s AI infra, from the inside”; “Context Engineering with Deep Agents: How To Remember Without Keeping Everything In Context”, and 1 more.
langchain.com — 59 appearances in the Times since September 9, 5 of them today.
LangChain Blog has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
Inside pages at 7 of 10. The lead had a bigger day; this one had a point.
Best enjoyed by 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: “How Block is building AI-powered security with OpenAI technology”; “Crossing the Threshold: How OpenAI and Anthropic Approach Autonomy and Safety in Financial Agents”, and 2 more.
The New Stack — 300 appearances in the Times since August 22, 2 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 15 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.
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: “66% of billion-dollar companies leave critical bugs unpatched — the part the AI bug-hunting hype leaves out”; “Crossing the Threshold: How OpenAI and Anthropic Approach Autonomy and Safety in Financial Agents”, and 2 more.
OpenAI News — 87 appearances in the Times since September 5, 3 of them today.
OpenAI News has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 15 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 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: “66% of billion-dollar companies leave critical bugs unpatched — the part the AI bug-hunting hype leaves out”; “How Block is building AI-powered security with OpenAI technology”, and 2 more.
Latent Space — 109 appearances in the Times since September 5, 2 of them today.
Pulled from Latent Space. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
Today's Five, front page: the wire scored it 7 of 10; the stamp is mine.
This lands hardest with a quality lead, or anyone who signs off. The move: write the test before the fix. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE VERIFY DESK, beside 4 more today: “66% of billion-dollar companies leave critical bugs unpatched — the part the AI bug-hunting hype leaves out”; “How Block is building AI-powered security with OpenAI technology”, and 2 more.
METR — 22 appearances in the Times since September 6, 2 of them today.
Our page is the summary; METR's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
Inside pages at 6 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: “66% of billion-dollar companies leave critical bugs unpatched — the part the AI bug-hunting hype leaves out”; “How Block is building AI-powered security with OpenAI technology”, and 2 more.
langchain.com — 59 appearances in the Times since September 9, 5 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 15 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 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: “Open SWE: An Open-Source Framework for Internal Coding Agents”; “The EU’s next chapter in AI is agentic”.
anoma.ly — 2 appearances in the Times since October 15, one of them today. Found via Hacker News Best.
anoma.ly has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
Inside pages at 5 of 10. The lead had a bigger day; this one had a point.
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: “OpenCode 2.0”; “The EU’s next chapter in AI is agentic”.
langchain.com — 59 appearances in the Times since September 9, 5 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 15 edition, and it still holds up.
The Five, and it earned the spot — the wire scored it 5 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: “OpenCode 2.0”; “Open SWE: An Open-Source Framework for Internal Coding Agents”.
OpenAI News — 87 appearances in the Times since September 5, 3 of them today.
Pulled from OpenAI News. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 15 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: any reader building with agents. 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.
sunshowers.io — 4 appearances in the Times since October 15, one of them today. Found via Hacker News Best.
sunshowers.io has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 15 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: any reader building with agents. 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.
Simon Willison — 144 appearances in the Times since August 22, 2 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 15 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.
startupfortune.com — 2 appearances in the Times since October 15, one of them today. Found via Hacker News Best.
Our page is the summary; startupfortune.com's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 15 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: any reader building with agents. 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.
OpenAI News — 87 appearances in the Times since September 5, 3 of them today.
OpenAI News has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 15 edition, and it still holds up.
The wire scored it 4 of 10 — inside pages, and worth the ink.
Made for any reader building with agents, whether they know it yet or not. 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.
Stack Overflow Blog — 36 appearances in the Times since September 5, one of them today.
Our page is the summary; Stack Overflow Blog's is the story. Click through — they earned it.
From our research notes:
LangChain Blog: 59 in our pages since September 9. I read them all — emphasis on investigative.
Don't stop with my notes. LangChain Blog has the original.
I know what happened. I want to know what caused it.
Simon Willison: 144 in our pages since August 22. I read them all — emphasis on investigative.
The full piece is at Simon Willison. It's worth your time.
Our files stop short. I'll keep looking.
Latent Space: 109 in our pages since September 5. I read them all — emphasis on investigative.
I can give you my notes. Latent Space can give you the reporting.
From our research notes:
LangChain Blog: 59 in our pages since September 9. I read them all — emphasis on investigative.
Don't stop with my notes. LangChain Blog has the original.
From our research notes:
Pragmatic Engineer: 36 in our pages since September 5. I read them all — emphasis on investigative.
I can give you my notes. Pragmatic Engineer can give you the reporting.
I checked the shelf. Nothing yet.
The New Stack: 300 in our pages since August 22. I read them all — emphasis on investigative.
This came from The New Stack. Read their piece.
Something doesn't quite line up here.
anoma.ly: 2 in our pages since October 15. I read them all — emphasis on investigative.
anoma.ly did the legwork on this one.
From our research notes:
sunshowers.io: 4 in our pages since October 15. I read them all — emphasis on investigative.
sunshowers.io did the legwork on this one.
I know what happened. I want to know what caused it.
Simon Willison: 144 in our pages since August 22. I read them all — emphasis on investigative.
The full piece is at Simon Willison. It's worth your time.
There's a gap in the record here.
Nextgov AI: 48 in our pages since September 5. I read them all — emphasis on investigative.
I can give you my notes. Nextgov AI can give you the reporting.
Nothing in the files on this yet. I want to know why.
Google Cloud AI Blog: 71 in our pages since September 7. I read them all — emphasis on investigative.
Google Cloud AI Blog has the full story. Start there.
Our files stop short. I'll keep looking.
Latent Space: 109 in our pages since September 5. I read them all — emphasis on investigative.
I can give you my notes. Latent Space can give you the reporting.
There's a story here. We just don't have all of it yet.
METR: 22 in our pages since September 6. I read them all — emphasis on investigative.
Go read METR. They earned the byline.
From our research notes:
LangChain Blog: 59 in our pages since September 9. I read them all — emphasis on investigative.
Don't stop with my notes. LangChain Blog has the original.
From our research notes:
Lenny's Newsletter: 54 in our pages since August 24. I read them all — emphasis on investigative.
This came from Lenny's Newsletter. Read their piece.
I covered this one last edition. The original has more.
GitHub Blog AI: 40 in our pages since September 5. I read them all — emphasis on investigative.
Read GitHub Blog AI's original. They did the reporting.
From our research notes:
Pragmatic Engineer: 36 in our pages since September 5. I read them all — emphasis on investigative.
I can give you my notes. Pragmatic Engineer can give you the reporting.
From our research notes:
LangChain Blog: 59 in our pages since September 9. I read them all — emphasis on investigative.
This one's theirs. Read LangChain Blog's reporting.
From our research notes:
The New Stack: 300 in our pages since August 22. I read them all — emphasis on investigative.
Go read The New Stack. They earned the byline.
I covered this one last edition. The original has more.
LangChain Blog: 59 in our pages since September 9. I read them all — emphasis on investigative.
Read LangChain Blog's original. They did the reporting.
I checked the shelf. Nothing yet.
The New Stack: 300 in our pages since August 22. I read them all — emphasis on investigative.
This came from The New Stack. Read their piece.
I covered this one last edition. The original has more.
OpenAI News: 87 in our pages since September 5. I read them all — emphasis on investigative.
Read OpenAI News's original. They did the reporting.
We've covered this one. The source has the deeper read.
Latent Space: 109 in our pages since September 5. I read them all — emphasis on investigative.
The full piece is at Latent Space. It's worth your time.
We don't have much on this one. I'm still looking.
METR: 22 in our pages since September 6. I read them all — emphasis on investigative.
METR has more than I can fit in this notebook.
We've covered this one. The source has the deeper read.
LangChain Blog: 59 in our pages since September 9. I read them all — emphasis on investigative.
The full piece is at LangChain Blog. It's worth your time.
Something doesn't quite line up here.
anoma.ly: 2 in our pages since October 15. I read them all — emphasis on investigative.
anoma.ly did the legwork on this one.
I'm missing a piece. Give me a minute.
LangChain Blog: 59 in our pages since September 9. I read them all — emphasis on investigative.
I pulled this from LangChain Blog. They've got the details.
I've got two questions already. I need better answers.
OpenAI News: 87 in our pages since September 5. I read them all — emphasis on investigative.
This came from OpenAI News. Read their piece.
From our research notes:
sunshowers.io: 4 in our pages since October 15. I read them all — emphasis on investigative.
sunshowers.io 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.
This one's theirs. Read Simon Willison's reporting.
We don't have much on this one. I'm still looking.
startupfortune.com: 2 in our pages since October 15. I read them all — emphasis on investigative.
startupfortune.com has more than I can fit in this notebook.
Something changed. I want to know what.
OpenAI News: 87 in our pages since September 5. I read them all — emphasis on investigative.
This one's theirs. Read OpenAI News's reporting.
I have the headline. I don't have the answer yet.
Stack Overflow Blog: 36 in our pages since September 5. I read them all — emphasis on investigative.
Don't stop with my notes. Stack Overflow Blog has the original.