Back issues
Josh Bersin's May analysis of ServiceNow's announcements: an 'Action Fabric' delivered as an MCP server that authenticates, permission-scopes, audits and monitors any agent from any vendor, built on the CMDB and workflow rules ServiceNow has spent twenty years assembling; Moveworks rebranded as 'Otto,' the front door for employees; and a demo of a rogue agent detected and neutralized from signals across four security vendors. Bersin calls it a 'tollgate for agents' and lines it up against Microsoft's Agent 365 and Workday's Agent System of Record making the same bet. His sharpest line is economic: if you buy AI to replace payroll with token fees and the agents are not doing more, the governance spend may not pay. Wrapped in a wall of site navigation but the prose is there. Published May 6, 2026; a late arrival; truncated at the cap.
Recap of the leftover questions from a live AMA with the four authors of Observability Engineering, second edition. The AI answers are the reason to read it: George describes copilot, commander and caretaker as agent personas that differ only in whether the human gates before or after the action, and says most teams should start at copilot; Austin explains that a wrong AI answer can come from the model, the prompt, the data and the infrastructure all at once, so attribution has to be baked into traces at the client; Liz tells a young engineer to specialise in what Claude cannot do, longer-horizon learning and judgment about wider design, rather than the writing of code; Charity's advice for colleagues who answer everything with 'Claude says' is to ask for their own opinion, in good humour. Worth a desk as leaders talking candidly; published a month ago.
OpenAI's customer story on 1Password, with the CTO Nancy Wang quoted throughout. The numbers are the customer's, published by the vendor: a 20.9% productivity improvement and a 10.9% cut in median pull-request cycle time for the core cohort, a multi-service production investigation dropping from about two hours to 5 to 20 minutes, and a modeled $783,750 of annual capacity that the piece unusually shows its arithmetic for: 50 developers times $250,000 fully loaded, times 20.9%, times a 40% attribution to Codex, times 75% realization. That formula is the useful part for any leader trying to build the same case. Also worth a desk: 1Password keeps secret references rather than credentials in repos and resolves them at the point of action so plaintext never enters model context, and has packaged its security policies as reusable AppSec skills. The house counting its own shelves, but with the receipts laid out.
SAP's head of quantum, Carsten Polenz, arguing in a Fortune commentary that AI is commoditizing prediction, so the next advantage is decision architecture: modeling a quarter-end where accounts payable, receivables and sales each optimize locally and together produce the wrong enterprise outcome. He names a category, 'Enterprise Decision Computing,' says most of it is solvable classically today, and positions quantum as a later extension rather than the starting point. It is a leader-facing argument about what to build once intelligence is cheap, which touches the paper's thesis, but it is thin on evidence and is an SAP executive coining an SAP-shaped category. Published August 13, 2026.
OpenAI's business-facing launch post for GPT-6 Astra, a week after release. The house counts its own shelves, but the shelves have numbers: Terminal-Bench 4.0 at 57.9% versus 37.3% for GPT-5.6 Sol and 55.8% for Claude Fable 5.1, at what it estimates as 9% and 63% lower API cost per task; pricing of $10 per million input and $50 per million output tokens; and a claim the model was trained to finish tasks in fewer tokens with fewer retries. Customer quotes relay their own measurements: CodeRabbit says it caught about 20% more bugs in review and more than doubled the catch rate on cross-file issues, Basis reports a 20% pass-rate gain on five-hour-plus workflows, Datacurve cites 74% on DeepSWE. All of it is vendor-curated, but the cost-per-task framing is the argument every buyer is about to have.
Ollama's June post on its MLX engine update: NVFP4 quantization that roughly halves the quality loss of 4-bit versus unquantized weights on Gemma 4 12B, faster time-to-first-token, and a selective snapshotting scheme that saves model state where conversations branch so sub-agents and retries do not recompute long prompts. Numbers are the vendor's own, on its own engine. The practical point for a reader is that the local-inference floor keeps rising, which changes the cost math for running agents on a laptop. Published June 11, 2026; a late arrival.
Ollama's own engineering post on multi-token prediction for Gemma 4 on Apple Silicon: a small draft model proposes several tokens, the main model verifies them in one pass, and Ollama auto-tunes the draft length at runtime so speculation never makes things slower. The headline number is nearly 90% faster generation measured on the Aider polyglot coding-agent benchmark, with a fused kernel they contributed to MLX making the large matrix multiplies 2 to 2.5x faster. It is the house counting its own shelves, but the measurement is on a real agent workload and they say why synthetic benchmarks would mislead. Worth a desk because it is a clear primary-source explanation of why coding agents get cheaper and faster on local hardware. Published June 29, 2026; a late arrival.
A short Ollama 0.30 release note: up to 20% faster on NVIDIA hardware via llama.cpp optimizations, Vulkan on by default so AMD and Intel GPUs work without vendor drivers, and wider GGUF model compatibility with a walkthrough for running a downloaded GGUF file inside coding agents. True and relevant but thin; a changelog with one number. Published June 5, 2026; a late arrival.
Ollama's note that OpenJarvis 1.0, an open-source local-first personal agent framework from Stanford's Hazy Research and Scaling Intelligence labs, now runs with Ollama. The framework tracks energy, cost and latency alongside accuracy as part of an 'intelligence per watt' research line, and ships presets for a morning digest agent, local document research, and a coding agent. Mostly install instructions, but the idea that a personal agent should default to local and measure its own watts is on thesis. Published May 28, 2026; a late arrival.
AINews on OpenAI's claimed Navier-Stokes result: the headline says roughly 10,000 agents, 130B tokens, more than $40M, 88 hours. The body is more careful than the headline. The most concrete public claim is one OpenAI researcher's tweet that the solution came from about 10,000 agents trained for a year with multi-agent RL to organize themselves under 'huge amounts of unstructured parallel test-time compute.' The newsletter then lists what the tweets do not establish: no theorem statement, no preprint, no verification artifact, and the '88 hours' figure appears only in a satirical post. A room would stop for it because, true or not, it is the clearest articulation yet of the bet that coordination and search at inference time matter as much as a bigger single model. Paid post; body truncated at the cap. All figures are the newsletter's reading of tweets, not a primary source.
Together AI's long explainer of the 'MIGHT' stack for agentic development on open models: model, inference provider, gateways and routers, harness, and tools (skills and MCP), with the argument that keeping the layers independent lets you swap models in minutes rather than rebuild a workflow. It walks through model sizes (Kimi K3 at 1.8T total parameters as the large example), where the harness sits, how MCP servers plug in, and starts a section on managing context before the body hits the capture limit. A vendor writing about the stack it sells, but the framing is genuinely useful for anyone who only knows Claude Code and wants to understand what is underneath. Truncated at the cap; scored on what was readable.
A late arrival from June 23, 2026: Mistral's launch post for OCR 4, positioned as the ingestion component for enterprise search and RAG pipelines. What is new is structure rather than text: bounding boxes, typed block classification for tables, equations and signatures, and inline confidence scores meant to drive source-grounded citations, redactions and human-in-the-loop checks. It claims a 72% average win rate in human preference tests and 85.20 on OlmOCRBench, and to its credit spends a section on the known limits of those benchmarks. Priced at $4 per thousand pages. Retrieval plumbing from the house, thin for the paper but real.
Nielsen Norman Group's Tony Alicea argues that as more interface work is AI-generated, the output of UX research and design shifts from documents written for humans to curated context that steers what the AI produces, because a model trained on every search screen will otherwise design the average one. He points to Google Labs' DESIGN.md (open-sourced April 2026, a file beside the code with exact tokens plus prose on how to apply them) and proposes a broader UX.md: research findings stated as constraints, interaction standards, a glossary in the users' own words, user models and 'world models' of the conditions of use, possibly as an index to a folder or served via MCP and skills. The frame he lands on is curation, not handoff: everyone in the organisation is now designing through prompts, and the job is making sure the organisation's knowledge is in the context whoever types. A named research firm making the prompts-as-code argument for the design discipline, dated July 24.
The Register's Simon Sharwood on a Microsoft Edge team post: AI-assisted coding has pushed a flood of new browser extensions into the Edge Add-ons review pipeline, turnaround times have risen, and the team, thinner after redundancies, is answering by automating the repeatable validation checks so human reviewers spend time on 'complex cases.' The Register notes this is the second Microsoft team in weeks (after Exchange, which said AI found so many bugs it could not ship an update) to say AI output is outrunning its capacity to check it. Secondhand from a vendor blog post, but the pattern is the point: generation got cheap and checking did not.
Harrison Chase's June 2024 post on why LLM-as-a-judge works and how LangSmith made judges self-improving. The argument is still clean: judging is easier than generating, and the judge often has information the generator lacked, such as the reference answer, so an LLM can grade RAG correctness, hallucination and toxicity where hard-coded rules cannot. The catch is that the judge prompt needs its own round of prompt engineering; their fix stores human corrections to the judge's verdicts as few-shot examples fed back into the judge, citing Shreya Shankar's 'Who Validates the Validators' as the motivating paper. A vendor explaining its own feature, and two years old, but as a plain explanation of the evaluation problem it holds up.
Dan Maccarone's long essay reading Shakespeare's magic as a model of AI output: the witches in Macbeth never lie, they equivocate, and Macbeth supplies the certainty himself; Malvolio finds his own name in four random letters, which is apophenia, the flattery a chatbot offers on demand; Hamlet at least staged the play to test the ghost, while we do not think to check. He ties it to the 'b*******' paper by Hicks, Humphries and Slater, and to the 1606 equivocation trials. The practical point, that a plausible answer that is technically true still needs a human to verify it and that the failure is usually the reader's, is sound, and it is memorable; it is a literary argument rather than a method. Cut off by the fetch cap before Prospero drowning his book gets its turn.
A Honeycomb guide to AI model drift: a model's quality quietly declines while uptime and latency stay green, because the data, users or business have moved since it was evaluated. It sorts drift into categories (data drift, and shifts from prompts, retrieval, embeddings and tool calls that change even when the model does not) and argues that observability across prompts, model calls, evaluations and token usage is how teams catch it. It is a getting-started explainer that ends in a product section on how Honeycomb helps. True and relevant to the verification desk, but thin and vendor-shaped; published a month ago.
OpenAI's Chief Global Affairs Officer Chris Lehane lays out the company's policy position after the GPT-6 Astra launch: it wants Congress to pass mandatory, capability-based national AI safety rules, will back four named California bills in the meantime, and will build voluntary frontier-lab standards with or without government. The one concrete disclosure is internal: for Astra, OpenAI says it added universal monitoring of full trajectories including chains of thought and a mandatory alignment-evaluation gate before broader internal deployment. It is a primary source about its own posture, written as advocacy. Policy rather than practice, so it stays on the wire.
Josh Bersin's May analysis of ServiceNow's announcements: an 'Action Fabric' delivered as an MCP server that authenticates, permission-scopes, audits and monitors any agent from any vendor, built on the CMDB and workflow rules ServiceNow has spent twenty years assembling; Moveworks rebranded as 'Otto,' the front door for employees; and a demo of a rogue agent detected and neutralized from signals across four security vendors. Bersin calls it a 'tollgate for agents' and lines it up against Microsoft's Agent 365 and Workday's Agent System of Record making the same bet. His sharpest line is economic: if you buy AI to replace payroll with token fees and the agents are not doing more, the governance spend may not pay. Wrapped in a wall of site navigation but the prose is there. Published May 6, 2026; a late arrival; truncated at the cap.
A late arrival from June 24, 2026: Mistral's engineering post on connector controls for agents in production. The premise is the right one, that connecting an agent to CRM, repos and inboxes is easy in a demo and stalls in production, and the answer is administrative: per-workspace and per-org connector access, tool-level switches that can block a single action like delete_file while the connector stays connected, API keys scoped to connectors so automated jobs never impersonate the person who wrote them, multi-account connectors, and a debugger for root-causing broken MCP connections. Over 60 pre-built connectors plus custom MCP. A product post by the house, but it is a clear statement of the doorman problem and one vendor's shape for the answer.
GitHub's monthly availability report for August, a bad month by its own account, written by an engineering leader. It reports the first production MySQL primaries running from Azure, roughly a million queries per second removed from the oldest shared database, and a 33% job-routing shift that brought Actions cache CPU from 98% to 80% for about three months of headroom, which it calls containment, not a finish line. The August 6 incident is the one that touches the paper's readers: a routine deployment reduced capacity in one site, and Actions plus Copilot coding agent, Copilot code review, Dependabot and Pages stalled for about ten hours across at least 74 organizations. Primary, numeric and candid about the platform under agent load, but the body is cut at the fetch cap partway through the incident write-ups.
LangChain's public-beta announcement for LangSmith LLM Gateway, a governance layer between agents and model providers, dated July 30. It offers spend caps at organisation, workspace, API-key and user level (a 402 when hit), rate limits, model fallbacks when a provider is down, per-customer limits via a request header for multi-tenant setups, PII and secret redaction before calls reach providers (enterprise only), and gateway events logged into traces. The motivating scenarios are concrete: a provider outage taking a customer-service agent down, or a coding agent stuck in a retry loop overnight making 10,000 calls for a four-figure invoice. A product post with a partner quote from Fireworks; the useful residue is the list of controls a production fleet needs first.
Today's paper: 21 stories, 5 on the front page, 6 desks staffed.
The presses were quiet this morning — this is the September 10 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 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 3 more today: “Bringing more control over your connectors | Mistral AI”; “GitHub availability report: August 2026”, and 1 more.
Josh Bersin — 20 appearances in the Times since September 6, one of them today.
Our page is the summary; Josh Bersin's is the story. Click through — they earned it.
THE DIRECTING DESK today: 3 stories.
The presses were quiet this morning — this is the September 10 edition, and it still holds up.
The wire scored it 6 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: “1Password increases engineering productivity 21% with Codex”; “SAP Head of Quantum: AI is about to commoditize intelligence. Better decisions will be the next competitive advantage | Fortune”.
Honeycomb Blog — 34 appearances in the Times since September 9, 2 of them today.
Our page is the summary; Honeycomb Blog's is the story. Click through — they earned it.
THE AMPLIFIER DESK today: 6 stories.
The presses were quiet this morning — this is the September 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 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 5 more today: “Ollama's highest performance on Apple Silicon yet with MLX · Ollama Blog”; “Faster Gemma 4 on MLX with multi-token prediction · Ollama Blog”, and 3 more.
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 LOOP DESK today: 1 story.
The presses were quiet this morning — this is the September 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 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 — alone there today, which says something about the story.
Together AI Blog — 15 appearances in the Times since September 6, one of them today.
Pulled from Together AI Blog. The original is worth your click — the byline earned it.
THE CONTEXT DESK today: 2 stories.
The presses were quiet this morning — this is the September 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 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 one more today: “UX-Context Design: Using UX Knowledge to Inform AI-Generated Design”.
Mistral AI News — 13 appearances in the Times since September 9, 2 of them today.
Mistral AI News 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 September 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.
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: “Aligning LLM-as-a-Judge with Human Preferences”; “What Shakespeare got right about AI”, and 2 more.
The Register — 11 appearances in the Times since September 5, one of them today.
The Register has the full story; we hold the door open. Go read the original.
THE FLEET DESK today: 4 stories.
The presses were quiet this morning — this is the September 10 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 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 3 more today: “Bringing more control over your connectors | Mistral AI”; “GitHub availability report: August 2026”, and 1 more.
Josh Bersin — 20 appearances in the Times since September 6, one of them today.
Our page is the summary; Josh Bersin's is the story. Click through — they earned it.
The presses were quiet this morning — this is the September 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 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 5 more today: “Ollama's highest performance on Apple Silicon yet with MLX · Ollama Blog”; “Faster Gemma 4 on MLX with multi-token prediction · Ollama Blog”, and 3 more.
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 September 10 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 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 3 more today: “Bringing more control over your connectors | Mistral AI”; “GitHub availability report: August 2026”, and 1 more.
Josh Bersin — 20 appearances in the Times since September 6, one of them today.
Our page is the summary; Josh Bersin's is the story. Click through — they earned it.
The presses were quiet this morning — this is the September 10 edition, and it still holds up.
The wire scored it 6 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: “1Password increases engineering productivity 21% with Codex”; “SAP Head of Quantum: AI is about to commoditize intelligence. Better decisions will be the next competitive advantage | Fortune”.
Honeycomb Blog — 34 appearances in the Times since September 9, 2 of them today.
Our page is the summary; Honeycomb Blog's is the story. Click through — they earned it.
The presses were quiet this morning — this is the September 10 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 5 more today: “GPT-6 Astra: The next generation in intelligence for work”; “Faster Gemma 4 on MLX with multi-token prediction · Ollama Blog”, and 3 more.
Ollama Blog — 14 appearances in the Times since September 9, 4 of them today.
Pulled from Ollama Blog. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the September 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.
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: “Aligning LLM-as-a-Judge with Human Preferences”; “What Shakespeare got right about AI”, and 2 more.
The Register — 11 appearances in the Times since September 5, one of them today.
The Register has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the September 10 edition, and it still holds up.
One of today's Five — the wire scored it 7 of 10, and I stamped it.
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 2 more today: “AMA: More Answers From the Observability Engineering Authors”; “SAP Head of Quantum: AI is about to commoditize intelligence. Better decisions will be the next competitive advantage | Fortune”.
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 September 10 edition, and it still holds up.
The Five, and it earned the spot — the wire scored it 7 of 10, I chose it.
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 5 more today: “GPT-6 Astra: The next generation in intelligence for work”; “Ollama's highest performance on Apple Silicon yet with MLX · Ollama Blog”, and 3 more.
Ollama Blog — 14 appearances in the Times since September 9, 4 of them today.
Pulled from Ollama Blog. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the September 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.
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 5 more today: “GPT-6 Astra: The next generation in intelligence for work”; “Ollama's highest performance on Apple Silicon yet with MLX · Ollama Blog”, and 3 more.
Ollama Blog — 14 appearances in the Times since September 9, 4 of them today.
Ollama Blog has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the September 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 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 3 more today: “ServiceNow Bets Big on Enterprise AI With Vision of Managing Everything JOSH BERSIN”; “GitHub availability report: August 2026”, and 1 more.
Mistral AI News — 13 appearances in the Times since September 9, 2 of them today.
Our page is the summary; Mistral AI News's is the story. Click through — they earned it.
The presses were quiet this morning — this is the September 10 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 3 more today: “ServiceNow Bets Big on Enterprise AI With Vision of Managing Everything JOSH BERSIN”; “Bringing more control over your connectors | Mistral AI”, and 1 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 September 10 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 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 5 more today: “GPT-6 Astra: The next generation in intelligence for work”; “Ollama's highest performance on Apple Silicon yet with MLX · Ollama Blog”, and 3 more.
Ollama Blog — 14 appearances in the Times since September 9, 4 of them today.
Pulled from Ollama Blog. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the September 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 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 3 more today: “ServiceNow Bets Big on Enterprise AI With Vision of Managing Everything JOSH BERSIN”; “Bringing more control over your connectors | Mistral AI”, and 1 more.
langchain.com — 59 appearances in the Times since September 9, 2 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 September 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 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 one more today: “UX-Context Design: Using UX Knowledge to Inform AI-Generated Design”.
Mistral AI News — 13 appearances in the Times since September 9, 2 of them today.
Mistral AI News has the full story; we hold the door open. Go read the original.
The presses were quiet this morning — this is the September 10 edition, and it still holds up.
The wire scored it 7 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 one more today: “Mistral OCR 4 : SOTA OCR for Document Intelligence”.
Nielsen Norman Group — 26 appearances in the Times since September 5, one of them today.
Our page is the summary; Nielsen Norman Group's is the story. Click through — they earned it.
The presses were quiet this morning — this is the September 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 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 — alone there today, which says something about the story.
Together AI Blog — 15 appearances in the Times since September 6, one of them today.
Pulled from Together AI Blog. The original is worth your click — the byline earned it.
The presses were quiet this morning — this is the September 10 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 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: “Another Microsoft team admits it’s struggling to handle flood of AI-generated code”; “What Shakespeare got right about AI”, and 2 more.
langchain.com — 59 appearances in the Times since September 9, 2 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 September 10 edition, and it still holds up.
The wire scored it 5 of 10 — inside pages, and worth the ink.
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: “Another Microsoft team admits it’s struggling to handle flood of AI-generated code”; “Aligning LLM-as-a-Judge with Human Preferences”, and 2 more.
UX Collective — 13 appearances in the Times since September 6, one of them today.
Our page is the summary; UX Collective's is the story. Click through — they earned it.
The presses were quiet this morning — this is the September 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 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: “Another Microsoft team admits it’s struggling to handle flood of AI-generated code”; “Aligning LLM-as-a-Judge with Human Preferences”, and 2 more.
Honeycomb Blog — 34 appearances in the Times since September 9, 2 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 September 10 edition, and it still holds up.
The wire scored it 4 of 10 — inside pages, and worth the ink.
Made for a quality lead, or anyone who signs off, whether they know it yet or not. 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: “Another Microsoft team admits it’s struggling to handle flood of AI-generated code”; “Aligning LLM-as-a-Judge with Human Preferences”, and 2 more.
OpenAI News — 87 appearances in the Times since September 5, 3 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 September 10 edition, and it still holds up.
Today's Five, front page: the wire scored it 8 of 10; the stamp is mine.
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 5 more today: “GPT-6 Astra: The next generation in intelligence for work”; “Ollama's highest performance on Apple Silicon yet with MLX · Ollama Blog”, and 3 more.
Latent Space — 109 appearances in the Times since September 5, one of them today.
Our page is the summary; Latent Space's is the story. Click through — they earned it.
The presses were quiet this morning — this is the September 10 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: “AMA: More Answers From the Observability Engineering Authors”; “1Password increases engineering productivity 21% with Codex”.
Fortune Eye on AI — 108 appearances in the Times since September 8, one of them today.
Our page is the summary; Fortune Eye on AI's is the story. Click through — they earned it.
Nothing in the files on this yet. I want to know why.
Josh Bersin: 20 in our pages since September 6. I read them all — emphasis on investigative.
Josh Bersin has the full story. Start there.
You've seen my notes on this already. Go to the source for the rest.
Honeycomb Blog: 34 in our pages since September 9. I read them all — emphasis on investigative.
Honeycomb Blog has more than I can fit in this notebook.
Our files stop short. I'll keep looking.
OpenAI News: 87 in our pages since September 5. I read them all — emphasis on investigative.
I can give you my notes. OpenAI News can give you the reporting.
There's a gap in the record here.
Together AI Blog: 15 in our pages since September 6. I read them all — emphasis on investigative.
I can give you my notes. Together AI Blog can give you the reporting.
I'm missing a piece. Give me a minute.
Mistral AI News: 13 in our pages since September 9. I read them all — emphasis on investigative.
Read Mistral AI News's original. They did the reporting.
I covered this one last edition. The original has more.
The Register: 11 in our pages since September 5. I read them all — emphasis on investigative.
Read The Register's original. They did the reporting.
Nothing in the files on this yet. I want to know why.
Josh Bersin: 20 in our pages since September 6. I read them all — emphasis on investigative.
Josh Bersin has the full story. Start there.
Our files stop short. I'll keep looking.
OpenAI News: 87 in our pages since September 5. I read them all — emphasis on investigative.
I can give you my notes. OpenAI News can give you the reporting.
Nothing in the files on this yet. I want to know why.
Josh Bersin: 20 in our pages since September 6. I read them all — emphasis on investigative.
Josh Bersin has the full story. Start there.
You've seen my notes on this already. Go to the source for the rest.
Honeycomb Blog: 34 in our pages since September 9. I read them all — emphasis on investigative.
Honeycomb Blog has more than I can fit in this notebook.
I know what happened. I want to know what caused it.
Ollama Blog: 14 in our pages since September 9. I read them all — emphasis on investigative.
The full piece is at Ollama Blog. It's worth your time.
I covered this one last edition. The original has more.
The Register: 11 in our pages since September 5. I read them all — emphasis on investigative.
Read The Register'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.
Go read OpenAI News. They earned the byline.
The headline is ahead of our files on this one.
Ollama Blog: 14 in our pages since September 9. I read them all — emphasis on investigative.
I pulled this from Ollama Blog. They've got the details.
There's a story here. We just don't have all of it yet.
Ollama Blog: 14 in our pages since September 9. I read them all — emphasis on investigative.
Go read Ollama Blog. They earned the byline.
Something changed. I want to know what.
Mistral AI News: 13 in our pages since September 9. I read them all — emphasis on investigative.
Mistral AI News has the full story. Start there.
We don't have much on this one. I'm still looking.
GitHub Blog AI: 40 in our pages since September 5. I read them all — emphasis on investigative.
I can give you my notes. GitHub Blog AI can give you the reporting.
The headline is ahead of our files on this one.
Ollama Blog: 14 in our pages since September 9. I read them all — emphasis on investigative.
I pulled this from Ollama Blog. They've got the details.
There's a story here. We just don't have all of it yet.
LangChain Blog: 59 in our pages since September 9. I read them all — emphasis on investigative.
This came from LangChain Blog. Read their piece.
I'm missing a piece. Give me a minute.
Mistral AI News: 13 in our pages since September 9. I read them all — emphasis on investigative.
Read Mistral AI News's original. They did the reporting.
I've got two questions already. I need better answers.
Nielsen Norman Group: 26 in our pages since September 5. I read them all — emphasis on investigative.
Want the whole story? Nielsen Norman Group has it.
There's a gap in the record here.
Together AI Blog: 15 in our pages since September 6. I read them all — emphasis on investigative.
I can give you my notes. Together AI Blog can give you the reporting.
I know what happened. I want to know what caused it.
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.
Our files stop short. I'll keep looking.
UX Collective: 13 in our pages since September 6. I read them all — emphasis on investigative.
UX Collective has more than I can fit in this notebook.
Nothing in the files on this yet. I want to know why.
Honeycomb Blog: 34 in our pages since September 9. I read them all — emphasis on investigative.
This one's theirs. Read Honeycomb Blog's reporting.
Same story as last edition. I'd go back to the original.
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:
Latent Space: 109 in our pages since September 5. I read them all — emphasis on investigative.
Don't stop with my notes. Latent Space has the original.
Nothing in the files on this yet. I want to know why.
Fortune Eye on AI: 108 in our pages since September 8. I read them all — emphasis on investigative.
Fortune Eye on AI has the full story. Start there.