Back issues
Part 1 of the Stack Overflow production series: make your agents boring. Contain the model to one node so the rest is ordinary, testable code; an agent proposes a decision and never acts. Where a loop is unavoidable it gets a hard iteration cap, a per-capability tool allow-list and a full trace of every step. Written for regulated, high-consequence systems, with the contracts and types, not just the ideas.
Part 1 of the Stack Overflow production series: make your agents boring. Contain the model to one node so the rest is ordinary, testable code; an agent proposes a decision and never acts. Where a loop is unavoidable it gets a hard iteration cap, a per-capability tool allow-list and a full trace of every step. Written for regulated, high-consequence systems, with the contracts and types, not just the ideas.
Simon Willison's own read of Claude Haiku 5.5. The price matches GPT-6 Luna at $0.10/$0.50 per million tokens up to 100,000 tokens, then rises five times beyond that, where Luna is the better deal. His token counter shows the new tokenizer uses about 1.25 times as many tokens as Haiku 4.5 on the same prompt, which he calls a hidden price increase. Reasoning cannot be turned off and defaults to medium; his pelican test cost under a tenth of a cent at low effort and 3.4 cents at max. An independent pricing check that the launch post does not give you.
The State of Devs 2026 survey from Devographics, about 5,300 respondents, independent of any AI vendor. Nearly half had faced job insecurity, 62% report burnout at some point, and about two thirds felt less motivated recently. For the first time it asked about AI: sentiment split 49% positive and 42% negative, and nearly half disagreed that AI tools had a positive effect on their mental state, with write-ins on addictive prompting, weakened learning and overload from running agents in parallel. A real number on how the shift feels to the people inside it.
Jessica Wachtel at The New Stack ran GPT-6.1 Sol and GPT-6 Astra through three of her own tests five times each: CI triage, incident-log questions over 114,000 input tokens, and a dependency resolver graded by 120 hidden tests. Both models were perfect on all 15 runs; Sol cost $2.66 in total against Astra's $14.77 and was faster on two of the three tests. She puts the tokens, times and costs in a table. Fifteen runs is a sample, not a benchmark, and she says what the tests cover.
Anthropic's own launch post for Claude Haiku 5.5, pitched for high-volume work and as a subagent beside Opus 5.5 and Sonnet 5.5. The benchmark table is the house measuring its own model: GDPval-AA 1620 against 735 for Haiku 4.5 and 1437 for GPT-6 Luna, OSWorld 2.1 at 72.4%. It also halves Sonnet 5.5's cache-read price and adds a monthly API credit for Max and Team plans. The footnote concedes the new tokenizer uses slightly more tokens per task, the same point Simon Willison measured at about 1.25 times.
OpenAI's launch of a GPT-6 model for all ChatGPT users and of Intelligent UI, where answers arrive as interactive diagrams, forms and tools composed on the fly. Most of the body is the demo text of those interfaces (a bicycle breakdown, a cooking checklist), and the body stops at the fetch cap mid-demo. A vendor describing its own product with no measurement. Notable for the claim of 1.2 billion weekly users and for answers turning into small generated apps.
Part 2 of the Stack Overflow production series: evals as a deployment gate. If a prompt change can ship without an eval failing the build, you have a notebook, not evals; golden cases are curated like tests, every past failure included. After the gate, a shadow-eval loop watches for drift from model upgrades and prompt nudges that quietly hurt a slice of inputs. Gives the case format and the baseline mechanics.
Honeycomb's annotated OpenTelemetry Collector configuration for LLM applications: receive over OTLP, map OpenInference and OpenLLMetry attributes onto the GenAI semantic conventions, redact prompt and completion content before it leaves the network, and manage volume without sampling away whole conversations. A vendor how-to that ends at its own Agent Timeline, but the configuration is real and portable. Practical for anyone trying to see what their agents did. The body stops near the fetch cap.
At Microsoft's Windows event, reported by Fortune, Satya Nadella called Windows a hybrid intelligence system mixing cloud and local AI, with agents built into the OS and a new Surface Laptop Ultra. Event coverage with quotes. Keeps as wire for the agents-in-the-OS line.
LangChain on revamping skills in Deep Agents: a skill is a folder with a SKILL.md, and only its name and description sit in context until the task needs it. New features bind tools to a skill so they load only when the skill is read, let an app pin a skill the user asked for, and reload changed skills mid-thread. They say enterprise registries now run to thousands of skills. A vendor post, but a clear statement of progressive disclosure as a context-budget technique.
OpenSearch veterans told The New Stack they are launching Infino, a single retrieval layer for agents: data in Parquet on object storage that agents search, filter and join directly instead of through a warehouse, a search engine and a vector store kept in sync. The CEO's point that an agent issues many small queries at once and every result costs context is a real one. Otherwise a launch told through the founder's quotes, with no customer numbers.
An arXiv paper by Alexander Bastounis and two co-authors arguing that checking an AI's autoformalisation in Lean gives no guarantee the original natural-language proof is right, with OpenAI's announced Navier-Stokes blow-up proof as the motivating case. Their claim is that resolving ambiguity in mathematical prose so the translation is faithful sits arbitrarily high in the Solvability Complexity Index, harder than the halting problem, and they give worked examples of AI mistranslations into Lean. A primary source on the exact week the math claims landed: a green check proves the formal statement, and someone still has to prove the formal statement is the one you meant.
GitHub's secret-scanning product lead with nine quarters of GitHub's own data. One in three pull requests now involves an agent, up from under one in ten a year ago. Screened pushes grew 2.84 times while pushes carrying credentials grew 2.59 times, so per-push prevalence did not rise, and developers overrode push blocks less often (6.63% down to 3.93%); the argument is that developers are outpaced, not careless. Manual revocation takes about 40 days on average and one in five takes over 90. The house measuring its own platform, and the post closes on GitHub's new classifier for unstructured secrets, but the numbers are primary and the claim is testable.
Robinson, OpenAI's former safety lead, gave The New Stack his first interview since leaving. His argument: capability no longer waits for a new pretraining run, because reasoning training can be layered onto an existing base and tools and integrations change what a model can do with no training at all, so the company is shipping new capability and risk every Tuesday. He contrasts this with the months of safety work talked up after GPT-4. One named insider's account, not a measurement, but it says plainly why safety testing tied to model releases misses most of the change.
Part 3 of a Stack Overflow series on running LLM systems in production: confidence. The argument is that a model's self-reported certainty is miscalibrated, so a usable confidence score has to be composed from independent signals, checked for calibration, backed by a second-model judge on high-stakes calls, and used to route the rest to humans in a way that catches errors. It gives the thresholds and an override table for tuning them by slice. Practical and concrete.
Faros AI answers System76's ban on AI-written code in COSMIC with its own data and its own story. Their Claude spend hit the org cap halfway through a month and cut everyone off at once, then it happened again on Codex. Their telemetry on 22,000 developers shows PR size up 71.8%, unreviewed merges up 76.3%, time in QA up 300.6%, monthly incidents up 125.4% even as incidents per PR improved. The vendor is measuring its customers to sell its own fix, and says the agentic-review result is correlational. The point that holds is that review is now the bottleneck and a ban and a spend cap both close the door on good work with the bad.
The Strands team (AWS's agent framework) releases Strands Decider 2B, an open small decision model that picks from fixed options or returns a score instead of generating text. The post is candid about the trade: faster and more reliable at a given size, always answers inside the allowed set, but worse at complex problems and useless for coding or chat. It places itself in a new class of decision models that arrived this month. Useful as the plainest explanation in the pool of why a routing or classification step in an agent system might not need a language model.
Latent Space interviews Stacklok's Craig McLuckie and Joe Beda on Mecatl, an open-source cloud-native agent harness. Their case is that desktop harnesses put the loop, local execution and session state in one process, while enterprise code and context are the IP that needs management, and they ask why agent identity is still modelled on human identity. Founders on their own project, no numbers. On thesis for anyone deciding where agents should run.
An aggregator rewrite of a report by The Information that Meta and Microsoft are cutting employee use of Claude in favour of their own coding tools. The figures are relayed second-hand: Microsoft's expected internal Anthropic spend of over $1 billion a year is said to be down by more than a third, and monthly AI spending limits in its cloud and AI group cut from $100,000 to about $10,000 per employee, which the piece correctly notes are ceilings, not actual spend. The chain is a small tech blog citing a paywalled outlet citing unnamed sources. The idea worth a desk is that per-seat agent budgets are now a management lever.
OpenAI's API documentation for the Decisions API in public beta: typed answers (a probability, a choice from a set, a rubric score) about ten times faster than the Responses API, on gpt-6-luna only. A primary source for how the request is shaped, three question types and a named answers array. The body is reference docs and runs into SDK code before stopping at the fetch cap. Pair it with the Strands and Liquid decision-model rows for the week's pattern.
The New Stack on OpenAI's Decisions API going public on GPT-6 Luna, returning predicates, choices and scores, set against Perplexity, Cloudflare and Amazon releasing open-weight decision models the same month. OpenAI's demo has it choosing lanes in a driving game and steering a robot's gaze. Mostly a relay of OpenAI's own video; the speed claims are the vendor's.
MIT News obituary for Margaret Hamilton, who led the more than 400 people writing Apollo's onboard flight software and died at 90. The body covers her path from weather prediction code with Edward Lorenz to the SAGE air defense system, where she took up software reliability as a then-new concept, to being the Apollo project's first programmer at the Instrumentation Lab. Real reporting from her own institution, heavy with photo captions at the head and site menus at the tail. Worth a line on any desk that talks about software rigor, since she is the reason the phrase software engineering exists.
ByteByteGo walks through how Netflix built GenRec, an LLM fine-tuned on viewing history to score and rank titles. It covers why the old ranker with thousands of hand-built features was costly, the two training phases and reward-weighted training. The post states it is assembled from public sources. The address reads how-uber-built-a-genie-to-answer while the body and title are about Netflix, so the feed's link and the article do not match. Opens with conference ads and stops at the fetch cap.
Today's paper: 23 stories, 5 on the front page, 6 desks staffed.
The presses were quiet this morning — this is the October 8 edition, and it still holds up.
One of today's Five — the wire scored it 7 of 10, and I stamped it.
The reader this serves: an engineering lead who directs more than they type. 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 — alone there today, which says something about the story.
Stack Overflow Blog — 36 appearances in the Times since September 5, 3 of them today.
Stack Overflow Blog has the full story; we hold the door open. Go read the original.
THE DIRECTING DESK today: 1 story.
The presses were quiet this morning — this is the October 8 edition, and it still holds up.
One of today's Five — the wire scored it 7 of 10, and I stamped it.
The reader this serves: an engineering lead who directs more than they type. 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 — alone there today, which says something about the story.
Stack Overflow Blog — 36 appearances in the Times since September 5, 3 of them today.
Stack Overflow Blog has the full story; we hold the door open. Go read the original.
THE AMPLIFIER DESK today: 5 stories.
The presses were quiet this morning — this is the October 8 edition, and it still holds up.
The wire scored it 7 of 10 — inside pages, and worth the ink.
This lands hardest with anyone whose day holds more work than hours. The move: try it on the one task that ate yesterday. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE AMPLIFIER DESK, beside 4 more today: “State of Devs 2026”; “GPT-6.1 Sol vs. GPT-6 Astra: Same accuracy at 18% of the cost”, and 2 more.
Simon Willison — 144 appearances in the Times since August 22, one of them today.
Our page is the summary; Simon Willison's is the story. Click through — they earned it.
THE LOOP DESK today: 3 stories.
The presses were quiet this morning — this is the October 8 edition, and it still holds up.
The wire scored it 7 of 10 — inside pages, and worth the ink.
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: “OpenTelemetry Collector Configuration for LLM Observability”; “Microsoft shows off new Windows software, revamped for agentic AI | Fortune”.
Stack Overflow Blog — 36 appearances in the Times since September 5, 3 of them today.
Our page is the summary; Stack Overflow Blog's is the story. Click through — they earned it.
THE CONTEXT DESK today: 2 stories.
The presses were quiet this morning — this is the October 8 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 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 one more today: “OpenSearch veterans launch Infino. Here's why it matters for agent builders.”.
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 VERIFY DESK today: 5 stories.
The presses were quiet this morning — this is the October 8 edition, and it still holds up.
Today's Five, front page: the wire scored it 8 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: “Secret protection must scale with software”; “OpenAI’s former safety lead says it’s shipping "new capability and risk every Tuesday"”, and 2 more.
arxiv.org — 8 appearances in the Times since September 20, one of them today. Found via Hacker News Best.
Our page is the summary; arxiv.org's is the story. Click through — they earned it.
THE FLEET DESK today: 5 stories.
The presses were quiet this morning — this is the October 8 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 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 4 more today: “Can a Cloud-Native Harness Make Agents Reliable Beyond the Desktop?”; “Meta and Microsoft Limit Employee Use of Claude AI Tools”, and 2 more.
strandsagents.com — 2 appearances in the Times since October 8, one of them today. Found via Hacker News Best.
Pulled from strandsagents.com. The original is worth your click — the byline earned it.
THE WIRE today: 2 stories.
The presses were quiet this morning — this is the October 8 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.
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.
news.mit.edu — 2 appearances in the Times since October 8, one of them today. Found via Hacker News Best.
news.mit.edu has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 8 edition, and it still holds up.
One of today's Five — the wire scored it 7 of 10, and I stamped it.
The reader this serves: an engineering lead who directs more than they type. 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 — alone there today, which says something about the story.
Stack Overflow Blog — 36 appearances in the Times since September 5, 3 of them today.
Stack Overflow Blog has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 8 edition, and it still holds up.
The wire scored it 7 of 10 — inside pages, and worth the ink.
This lands hardest with anyone whose day holds more work than hours. The move: try it on the one task that ate yesterday. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE AMPLIFIER DESK, beside 4 more today: “State of Devs 2026”; “GPT-6.1 Sol vs. GPT-6 Astra: Same accuracy at 18% of the cost”, and 2 more.
Simon Willison — 144 appearances in the Times since August 22, one 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 8 edition, and it still holds up.
One of today's Five — the wire scored it 7 of 10, and I stamped it.
The reader this serves: anyone whose day holds more work than hours. 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 4 more today: “Claude Haiku 5.5”; “GPT-6.1 Sol vs. GPT-6 Astra: Same accuracy at 18% of the cost”, and 2 more.
2026.stateofdevs.com — 2 appearances in the Times since October 8, one of them today. Found via Hacker News Best.
2026.stateofdevs.com has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 8 edition, and it still holds up.
One of today's Five — the wire scored it 7 of 10, and I stamped it.
The reader this serves: anyone whose day holds more work than hours. 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 4 more today: “Claude Haiku 5.5”; “State of Devs 2026”, and 2 more.
The New Stack — 300 appearances in the Times since August 22, 4 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 8 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.
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 4 more today: “Claude Haiku 5.5”; “State of Devs 2026”, and 2 more.
Anthropic Blog — 10 appearances in the Times since September 5, one of them today. Found via Hacker News Best.
Anthropic Blog has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 8 edition, and it still holds up.
The wire scored it 5 of 10 — inside pages, and worth the ink.
Made for anyone whose day holds more work than hours, whether they know it yet or not. The move: try it on the one task that ate yesterday. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE AMPLIFIER DESK, beside 4 more today: “Claude Haiku 5.5”; “State of Devs 2026”, and 2 more.
OpenAI News — 87 appearances in the Times since September 5, one of them today. Found via Hacker News Best.
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 8 edition, and it still holds up.
The wire scored it 7 of 10 — inside pages, and worth the ink.
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: “OpenTelemetry Collector Configuration for LLM Observability”; “Microsoft shows off new Windows software, revamped for agentic AI | Fortune”.
Stack Overflow Blog — 36 appearances in the Times since September 5, 3 of them today.
Our page is the summary; Stack Overflow Blog's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 8 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.
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: “Part 2: Evals as a deployment gate — and how to know when they drift - Stack Overflow”; “Microsoft shows off new Windows software, revamped for agentic AI | Fortune”.
Honeycomb Blog — 34 appearances in the Times since September 9, one of them today.
Honeycomb Blog has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 8 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 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: “Part 2: Evals as a deployment gate — and how to know when they drift - Stack Overflow”; “OpenTelemetry Collector Configuration for LLM Observability”.
Fortune Eye on AI — 108 appearances in the Times since September 8, one of them today.
Pulled from Fortune Eye on AI. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 8 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 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 one more today: “OpenSearch veterans launch Infino. Here's why it matters for agent builders.”.
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 8 edition, and it still holds up.
The wire scored it 5 of 10 — inside pages, and worth the ink.
This lands hardest with the one who feeds the agents' memory. 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 one more today: “Revamping Skills in Deep Agents”.
The New Stack — 300 appearances in the Times since August 22, 4 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 8 edition, and it still holds up.
Today's Five, front page: the wire scored it 8 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: “Secret protection must scale with software”; “OpenAI’s former safety lead says it’s shipping "new capability and risk every Tuesday"”, and 2 more.
arxiv.org — 8 appearances in the Times since September 20, one of them today. Found via Hacker News Best.
Our page is the summary; arxiv.org's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 8 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 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: “Navier-Stokes lost in translation: Why Lean verification of AI autoformalisation does not guarantee correct natural language proofs”; “OpenAI’s former safety lead says it’s shipping "new capability and risk every Tuesday"”, and 2 more.
GitHub Blog AI — 40 appearances in the Times since September 5, one of them today.
Pulled from GitHub Blog AI. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 8 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 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: “Navier-Stokes lost in translation: Why Lean verification of AI autoformalisation does not guarantee correct natural language proofs”; “Secret protection must scale with software”, and 2 more.
The New Stack — 300 appearances in the Times since August 22, 4 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 8 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: “Navier-Stokes lost in translation: Why Lean verification of AI autoformalisation does not guarantee correct natural language proofs”; “Secret protection must scale with software”, and 2 more.
Stack Overflow Blog — 36 appearances in the Times since September 5, 3 of them today.
Stack Overflow Blog has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 8 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: “Navier-Stokes lost in translation: Why Lean verification of AI autoformalisation does not guarantee correct natural language proofs”; “Secret protection must scale with software”, and 2 more.
Faros AI Blog — 47 appearances in the Times since September 5, one of them today.
Faros AI Blog has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 8 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 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 4 more today: “Can a Cloud-Native Harness Make Agents Reliable Beyond the Desktop?”; “Meta and Microsoft Limit Employee Use of Claude AI Tools”, and 2 more.
strandsagents.com — 2 appearances in the Times since October 8, one of them today. Found via Hacker News Best.
Pulled from strandsagents.com. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 8 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.
A leader scaling past one agent will get the most from this one. 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 4 more today: “Introducing Strands Decider 2B: a small, open source, decision model”; “Meta and Microsoft Limit Employee Use of Claude AI Tools”, and 2 more.
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.
The presses were quiet this morning — this is the October 8 edition, and it still holds up.
The wire scored it 5 of 10 — inside pages, and worth the ink.
This lands hardest with a leader scaling past one agent. The move: read it with the team lead in the room. No per-story read on file; this note is the desk speaking, not the story.
Filed on THE FLEET DESK, beside 4 more today: “Introducing Strands Decider 2B: a small, open source, decision model”; “Can a Cloud-Native Harness Make Agents Reliable Beyond the Desktop?”, and 2 more.
rswebsols.com — 2 appearances in the Times since October 8, one of them today. Found via Hacker News Best.
Our page is the summary; rswebsols.com's is the story. Click through — they earned it.
The presses were quiet this morning — this is the October 8 edition, and it still holds up.
Inside pages at 5 of 10. The lead had a bigger day; this one had a point.
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 4 more today: “Introducing Strands Decider 2B: a small, open source, decision model”; “Can a Cloud-Native Harness Make Agents Reliable Beyond the Desktop?”, and 2 more.
developers.openai.com — first appearance in the Times. Found via Hacker News Best.
Pulled from developers.openai.com. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the October 8 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.
A leader scaling past one agent will get the most from this one. 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 4 more today: “Introducing Strands Decider 2B: a small, open source, decision model”; “Can a Cloud-Native Harness Make Agents Reliable Beyond the Desktop?”, and 2 more.
The New Stack — 300 appearances in the Times since August 22, 4 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 8 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.
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.
news.mit.edu — 2 appearances in the Times since October 8, one of them today. Found via Hacker News Best.
news.mit.edu has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the October 8 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 any reader building with agents. The move: share it with whoever will ask about it first. 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.
ByteByteGo — 38 appearances in the Times since August 22, one of them today.
Pulled from ByteByteGo. The original is worth your click — the byline earned it.
Our files stop short. I'll keep looking.
Stack Overflow Blog: 36 in our pages since September 5. I read them all — emphasis on investigative.
Stack Overflow Blog did the legwork on this one.
Our files stop short. I'll keep looking.
Stack Overflow Blog: 36 in our pages since September 5. I read them all — emphasis on investigative.
Stack Overflow Blog did the legwork on this one.
Nothing in the files on this yet. I want to know why.
Simon Willison: 144 in our pages since August 22. I read them all — emphasis on investigative.
Simon Willison has the full story. Start there.
From our research notes:
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.
We don't have much on this one. I'm still looking.
LangChain Blog: 59 in our pages since September 9. I read them all — emphasis on investigative.
I can give you my notes. LangChain Blog can give you the reporting.
From our research notes:
arxiv.org: 8 in our pages since September 20. I read them all — emphasis on investigative.
arxiv.org has more than I can fit in this notebook.
The timing caught my eye. I'm checking that next.
strandsagents.com: 2 in our pages since October 8. I read them all — emphasis on investigative.
This came from strandsagents.com. Read their piece.
The timing caught my eye. I'm checking that next.
news.mit.edu: 2 in our pages since October 8. I read them all — emphasis on investigative.
Go read news.mit.edu. They earned the byline.
Our files stop short. I'll keep looking.
Stack Overflow Blog: 36 in our pages since September 5. I read them all — emphasis on investigative.
Stack Overflow Blog did the legwork on this one.
Nothing in the files on this yet. I want to know why.
Simon Willison: 144 in our pages since August 22. I read them all — emphasis on investigative.
Simon Willison has the full story. Start there.
There's a gap in the record here.
2026.stateofdevs.com: 2 in our pages since October 8. I read them all — emphasis on investigative.
2026.stateofdevs.com did the legwork on this one.
I know what happened. I want to know what caused it.
The New Stack: 300 in our pages since August 22. I read them all — emphasis on investigative.
This one's theirs. Read The New Stack's reporting.
From our research notes:
Anthropic Blog: 10 in our pages since September 5. I read them all — emphasis on investigative.
Read Anthropic Blog's original. They did the reporting.
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.
Want the whole story? OpenAI News has it.
From our research notes:
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.
There's a story here. We just don't have all of it yet.
Honeycomb Blog: 34 in our pages since September 9. I read them all — emphasis on investigative.
Go read Honeycomb Blog. They earned the byline.
Something changed. I want to know what.
Fortune Eye on AI: 108 in our pages since September 8. I read them all — emphasis on investigative.
The full piece is at Fortune Eye on AI. It's worth your time.
We don't have much on this one. I'm still looking.
LangChain Blog: 59 in our pages since September 9. I read them all — emphasis on investigative.
I can give you my notes. LangChain Blog can give you the reporting.
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.
From our research notes:
arxiv.org: 8 in our pages since September 20. I read them all — emphasis on investigative.
arxiv.org has more than I can fit in this notebook.
Something changed. I want to know what.
GitHub Blog AI: 40 in our pages since September 5. I read them all — emphasis on investigative.
The full piece is at GitHub Blog AI. 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.
I checked the shelf. Nothing yet.
Stack Overflow Blog: 36 in our pages since September 5. I read them all — emphasis on investigative.
Go read Stack Overflow Blog. They earned the byline.
I have the headline. I don't have the answer yet.
Faros AI Blog: 47 in our pages since September 5. I read them all — emphasis on investigative.
Read Faros AI Blog's original. They did the reporting.
The timing caught my eye. I'm checking that next.
strandsagents.com: 2 in our pages since October 8. I read them all — emphasis on investigative.
This came from strandsagents.com. Read their piece.
I covered this one last edition. The original has more.
Latent Space: 109 in our pages since September 5. I read them all — emphasis on investigative.
Read Latent Space's original. They did the reporting.
There's a gap in the record here.
rswebsols.com: 2 in our pages since October 8. I read them all — emphasis on investigative.
rswebsols.com has more than I can fit in this notebook.
We've covered this one. The source has the deeper read.
developers.openai.com: first time in our pages. I'll be reading them from here.
The full piece is at developers.openai.com. It's worth your time.
From our research notes:
The New Stack: 300 in our pages since August 22. I read them all — emphasis on investigative.
Read The New Stack's original. They did the reporting.
The timing caught my eye. I'm checking that next.
news.mit.edu: 2 in our pages since October 8. I read them all — emphasis on investigative.
Go read news.mit.edu. They earned the byline.
I have the headline. I don't have the answer yet.
ByteByteGo: 38 in our pages since August 22. I read them all — emphasis on investigative.
I pulled this from ByteByteGo. They've got the details.