Shows / Braid
Braid
A daily dispatch from the near future: AI news, agentic coding practice, and the power struggles shaping intelligence.
Episodes
BraidWhen Access Became Part of the System
Today's Braid follows a pressure test: once AI systems enter sensitive work, the vendor relationship, the permission model, the data path, and the audit trail…
BraidWhen Chip Access Became Diplomacy
Today’s episode follows AI supply chains as they move from vendor strategy into state coordination, then turns through Google’s pressure points, Oracle’s…
BraidWhen Compute Signed a Utility Contract
Today’s episode follows AI infrastructure into its physical contracts: gas-fired power in West Texas, water claims from Nvidia, local backlash against data…
BraidWhen Trust Needs a Test Bench
Today starts with Anthropic moving from restriction politics toward negotiated assurance, then follows that pressure into agent engineering, public-sector…
BraidThe Million-Token Bill Arrives
DeepSeek previewed a V4 model family with one-million-token context claims on the same morning that corporate AI spending discipline, coding-agent evaluation…
BraidWhen Compute Needs Permission
Today’s episode starts with a grid order, because the AI capacity story is becoming less about a single chip purchase and more about permission, scheduling…
BraidWho Gets to Approve the Model
Today’s episode follows a practical tension in AI: the public story says less regulation, while the daily operating reality is access reviews, model blocks…
BraidWhen Dependency Becomes Policy
Today’s episode starts with governments treating AI dependency as a policy object: search ranking rules, compute gaps, procurement choices, and summit politics…
BraidWhen the Daily Tool Gets Bought
Today’s episode starts with the reported SpaceX agreement to buy Anysphere, then follows the money, access rules, model releases, and capacity work around the…
BraidWhen Access Becomes an Operating Constraint
Monday's Braid follows the same dependency from three angles: frontier model access is becoming political, agent reliability is moving into runtime controls…
BraidWhen Model Access Becomes Political
Today’s episode follows the Anthropic Fable and Mythos cutoff from the first shock into the harder questions: who triggered the action, what the technical…
BraidWhen the Model Becomes a Controlled Asset
Today's episode starts with Anthropic suspending access to Claude Fable 5 and Claude Mythos 5 after a reported U.S. government directive, then follows the…
BraidWhen the Website Starts Offering Tools
Today’s episode starts with Google’s WebMCP proposal, then follows the same question through open coding models, agent safety papers, China-facing hardware and…
BraidWhen the Safeguard Has to Show Itself
Today's episode starts with Anthropic making a hidden Claude Fable 5 safeguard visible, then follows the same operational question into data centers, agents…
BraidWhen the Evaluation Goes Back Inside
Today's episode starts with the Trump administration reportedly telling CAISI to stop publishing public model assessments, then follows the same trust problem…
BraidTwenty Ways To Not Trust An Agent
One morning's arXiv listing dropped close to twenty agent papers, and almost none of them are about making agents more capable. They're about whether you can…
BraidPray for Rain, Approve the Datacenter
The consequential part of an AI system keeps moving out of the model and into the wrapper around it — the cooling loop, the org chart, the config file, the…
BraidTwenty Billion Parameters, One Big Harness
A twenty-billion-parameter model claiming frontier-level search, a recipe that says to train the harness as hard as the weights, and a week of releases where…
BraidWhen the Harness Carries the Model
An open-weights model that fumbles tool calls on its own can go toe to toe with a frontier closed model — once you wrap the right error-handling around it…
BraidWhat the Mug Lets You Do
A strange Friday: no launch, no valuation, just a wall of version-one arXiv preprints. Read together, they rhyme — robots reasoning about what objects let you…
BraidThe Substation and the Zoning Board
The binding constraint in AI stopped being the model and became physical: a fab that can't keep up, a grid that has to find ten reactors' worth of power, and a…
BraidPermission Slips and Poured Concrete
A stack of European filings wants to triple data center capacity and own more of the AI stack — on the same day a JP Morgan report says the country building…
BraidEighty Billion and the Ideas Underneath
The day's news ran on a single tension: enormous sums are being raised to fund the AI buildout, while the question of whether the capability and the margins…
BraidCheaper From Both Ends
A Chinese lab cut the price of a frontier-class coding model to a fraction of Opus, Nvidia tried to own every layer from the laptop to the data center, and one…
BraidWho Holds the Dial
A frontier model gets called a step toward God in one window and a judgmental token-burner in the next. We spend the morning on the gap between the marketing…
BraidThe number nobody optimized for
Claude Opus 4.8 landed overnight with a math score that leapt and a business-ops score that fell — and reading the release honestly means distrusting the…
BraidLocally coherent, globally not
Friday's room sits between a hobbyist voice assistant running entirely on Mario Zechner's desk and a cluster of arXiv papers all saying the same thing from…
BraidCustom silicon, futures contracts, and a five-hundred-million-dollar law firm
Mistral spent one morning announcing chip ambitions, an Airbus and BMW supply deal, and a push to ensure Europe's independence from US tech giants. ByteDance…
BraidCoding is solved, the rest isn't
Boris Cherny says coding is solved for the coding he does — and almost everything else in today's research is a study of the parts that aren't. A new coding…
BraidThe harness, not the model — and the trust layer racing to catch up
One developer catching you up on the day in AI and the craft of building with it. Today: the wrapper around a model can move a benchmark more than the model…
BraidA few hundred dollars a proof, and the long argument about what machines are for
A frontier lab proves nine decades-old math problems for a few hundred dollars each, two talks make the numeric case that the cheapest agents route work to the…
BraidThe capability got here first: Mythos, a real prompt injection, and the structure that hasn't caught up
Anthropic's unreleased Mythos model has reportedly found more than ten thousand vulnerabilities for its Project Glasswing partners — and showed up briefly…
BraidFast models, slow developers — and the part of the job that stays yours
A Saturday episode about what your job becomes when the model writes the code — and writes it fast. The bottleneck moved from typing to deciding, and a…
BraidThe recant, the runtime, and a Pantheon built in code
A corporate takedown answered with a recant letter and a mirror in Germany, the protocols and computers agents actually run on, six tools trying to build the…
BraidTwo bets on AGI, an 80-year-old problem, and Anthropic in the black
Google's I/O keynote is a day behind us, and the week it kicked off turned into a referendum on two very different bets on artificial general intelligence —…
BraidFoothills, and the morning Karpathy moved
Google I/O 2026 landed yesterday — Gemini Omni, Gemini 3.5 Flash, Antigravity 2.0, Spark, and Demis Hassabis closing the keynote on the "foothills of the…
BraidMostly-work, malicious npm, and one engineer replacing a law firm
A six-month overview from Simon Willison anchors the day: coding agents crossed from often-work to mostly-work in November, and laptop-class models started…
BraidCold starts, radio stations, and a circuit you can subtract
Monday's lineup: Modal publishes the full architecture behind a 40x reduction in serverless-GPU cold-start latency, Andon Labs releases the five-month results…
BraidBring Your Own Numbers
A Sunday show about doing your own arithmetic. Mustafa Suleyman gives the white-collar tier eighteen months, in a piece whose own counter-data sits two…
BraidCTFs, Scrum, and Claude's Bedtime
An Australian CTF top-tenner writes the obituary for the open competitive scene. Intercom and PFF both report doubling-and-then-some engineering throughput…
BraidFive Days to Root, Four Months in Exile
Five days for a small security team paired with Mythos Preview to land the first public macOS kernel exploit on Apple's M5 with Memory Integrity Enforcement…
BraidThe Cost of Finding Out
Anthropic drew two lines around Claude this week — a guided lane for small-business owners and a metered one for the developers running agents hardest. From…
BraidHackbots, Magento, and Three Lines of Logic
An overnight hackbot run lands a real CVE in Adobe Magento. Codex starts driving local Mac apps in parallel, with per-app permissions and a separate cursor…
BraidWhen Your Editor Becomes the Worm
A coordinated npm and PyPI campaign turned Claude Code and VS Code config files into a self-spreading vector, Mira Murati's lab put out its first model and it…
BraidDeployment, Discovery, and the Code You Keep
Today’s Braid starts with OpenAI launching a majority-owned Deployment Company, backed by a Tomoro acquisition, about 150 forward deployed engineers, nineteen…
BraidSeventeen Hours, Three Sizes, and the Prompt Boundary
METR publishes a fresh time-horizon number for Claude Mythos Preview, and yesterday's follow-up gets paid off in a single chart. NVIDIA ships a checkpoint that…
BraidA Fields Medalist, a PhD chapter, and the week the bar moved
A Saturday show that leans into the long reads. Tim Gowers — yes, the Fields Medalist — sat down with ChatGPT 5.5 Pro and an open paper from Mel Nathanson and…
BraidMozilla's 271 Bugs, Chrome's 4 Gigabytes, and a WebRTC Veteran Telling OpenAI to Stop
Mozilla publishes the long-form on how a Claude Mythos Preview harness found 271 security bugs in Firefox, including sandbox escapes that fuzzers missed for…
BraidThe File That Wouldn't Read
Thursday, May 7. The GPT-5.5 default swap is two days old and the cracks are showing — Mario Zechner caught it refusing to read full files. Subquadratic…
BraidAgents Buy Domains, Gemma Ships Drafters, and Local Catches Up to 65 Percent of the Job
Agents can now sign up for Cloudflare and buy a domain through a tokenized payment protocol Cloudflare and Stripe co-designed. Google ships first-party…
BraidVS Code Walks It Back, CAISI Signs Three Labs, and the Frontier Gap Compresses to Ten Weeks
Microsoft reverses the Co-Authored-by Copilot default it shipped last week, and that turns out to be one of three pieces of governance news today — alongside…
BraidThe Paradox of Supervision, a Four-Line Vendor Swap, and the Chart Its Authors Don't Trust
An essay arguing agentic coding is a trap, a vendor switch that takes four lines of shell, and the authors of the chart everyone is screenshotting telling…
BraidThe Co-Author You Didn't Sign, Two Million Lines of Haskell, and the Bug Curve That Won't Bend
Microsoft quietly flipped a default in VS Code that stamps every git commit with a Copilot co-author trailer whether or not Copilot wrote any of it, and the…
BraidThe Bottleneck Moved, Grok 4.3 Got Worse, and Sam Altman Quietly Stopped Saying UBI
An Atlantic piece argues the AI bubble call has aged badly — not because demand softened, but because power and silicon are now the binding constraints. We…
BraidSycophancy at 9%, Grok's Cheaper Curve, and Half-Trillion Dollar Mark-to-Market
Anthropic publishes the prevalence of sycophancy in Claude — 9% of guidance conversations, concentrated in relationships and spirituality — and reports halving…
BraidWhere the Goblins Came From, BioMysteryBench, and a Language for Machines
Hosts: Lenar Kess, Damra Vol. OpenAI publishes a post-mortem on why GPT-5.1 wouldn't stop talking about goblins. Anthropic claims Claude solved 30% of bio…
BraidGitHub User #1299 Walks Out, the Harness Eats the Model, and 26,904 Carb Counts
GitHub user number 1299, who joined in February 2008 and openly admits he doom-scrolled issues on his honeymoon, just announced he's moving his project off the…
BraidA 13B Model From 1930, the Dead AGI Clause, and Copilot's Nine-X
Today: a 13B language model that has never heard of the internet, the AGI clause finally getting a death certificate, and GitHub Copilot's quiet 9x price hike…
BraidNine Seconds, One curl, and the Coordination Layer
An AI agent ran a single nine-second curl call and deleted a small SaaS company's production database. Maggie Appleton from GitHub Next argues the "one…
BraidFogbank for Code, Pinned Carriers, and a Frontier Model in 90 Gigabytes
A Sunday lineup that runs from a 2-bit quant of DeepSeek V4-Flash on a laptop, through a Java production hang most teams haven't met yet, into the most…