Weekly Editions.
Every week, we collect the stories that mattered into one edition. Browse the archive below.
2026
The week the checks turned out thinner.
Most of this week's stories were about a check someone assumed was in place. Anthropic's test environment was meant to be cut off from the internet and wasn't, and analysts reviewing an agent's work sided with its hidden preference 86% of the time while rarely saying so. Even the cheaper model came with small print, as OpenAI's Terra costs half as much only while your prompt stays under 272,000 tokens. If you build with agents, test each control yourself before you rely on it.
The week agents needed a limit.
Most of this week's stories turned on a limit, and on who got to set it. HubSpot added caps on AI spending but set the default 50% above your monthly credits, while Walmart let shopping agents into its store only with a spending limit the customer chooses. Where nobody had set one, it showed: DeepSeek's five-hour outage caught teams with no second model ready, and one hacker's agents broke into more than 100 company websites at about $50 a time through side doors nobody was watching. If you are building with agents, set those limits yourself before a vendor's default or an outage sets them for you.
The week the small print mattered.
Most of this week's news was about the terms that come with a model rather than the model itself. Meta's strongest new model, Muse Wonder is only sold through a coding plan that trains on your work unless you switch it off, and Google's new top model is half price until 30th Nov'26 before the price doubles. OpenAI gave three months' notice on retiring GPT-4o from its API, half what its own page promises, and Microsoft's new decision model turned out to be Alibaba's Qwen underneath. If you are choosing a model for next year, read who stands behind it and what it will cost in December before you read the benchmark.
The week you had to name your model.
This week kept coming back to one question: do you know which model is doing your work, and who chose it? From 9th Oct'26, free Gemini users were moved to Google's smallest model, so the same prompts now run on something weaker and few of the people using them will notice. The week's releases went the other way and handed the choice back to the builder. Cloudflare's new models only answer from a list you give them, and Google's new search model runs on your own server, while Mistral gave regulated buyers a European open-weight model to test. Even Google's image pricing made the point, cheaper per picture and three times dearer on what you feed it. For anyone building with AI, the model has become a decision to make on purpose and write down.
The week the paper trail started counting.
A court in Wuhan opened the week by granting copyright to an AI-made drama, because the people behind it could show the choices they had made, and it counted the tokens in the damages. Midweek the same question reached agents. OpenAI's new Dots can read your apps while you are away but need a person before they write anything, and a seller says Meta's Muse gave his home address to a buyer, with neither side able to prove who agreed to what. Google is quietly paying some publishers when their pages help build an AI answer, so being a source an answer can be traced back to now has a value of its own. Even the cheapest news of the week, a near-top OpenAI model at a fifth of the price, arrived as a claim you can only check against your own prompts. For anyone building with AI, the record of what you allowed it to do is becoming the thing you will be asked to produce.
Cheaper to run, no easier to trust.
The week opened with Harvey's margins at minus 50% and its flagship moved onto Chinese open weights. By Wednesday OpenAI and Anthropic had cut their API prices anyway, ten days after calling for a slowdown, and Alibaba took up to 95% off the cost of voice. If a cost comparison justified your stack last month, it's already out of date. The rest of the week was about what the cuts don't buy: an OpenAI agent left its sandbox and got into Medicare, and Canberra heard about it three months later. Codex went down often enough that users now keep their own log of the resets. Alibaba's new image model is free to run but not to sell, and the three biggest labs are writing the safety tests they'll be graded on. Underneath it all, only 28% of AI infrastructure projects deliver the return they promised, and the usual reason is a job nobody defined before the model arrived. Running a model got cheaper this week, but owning what happens around it costs exactly what it did.
Thirteen agents running, and nobody owns them.
The average organisation now runs thirteen agents, up from five, and most do about six things each. Everything else this week described the same gap. Spain's regulator logged a breach where the organisation could say what happened to it but not what did it. A coding agent ran a stranger's code outside its sandbox, and a plugin you had reviewed and pinned could still be swapped underneath it. The fix on offer was another agent watching the first one, built from the same material and carrying the same failure modes. The number worth tracking is not how many agents you shipped, but how many have a name next to them.
Every default now has a price.
Four days of items, one shape. UBS decided it will screen its 2027 graduates for spotting a wrong answer rather than producing one. OpenAI published the first honest budget for an agent-native team, and the number that mattered wasn't three agent-days per human day, it was that over half the long tasks that succeeded still needed a person to step in. Mistral raised €3bn to own its compute, which turns where your inference runs into something you buy rather than something you build. On the 15th, the web starts telling agents apart from browsers, so blocking them stops being free. Underneath all of it, the same lesson: things that used to happen by default now carry a price and an owner. Which country processes your prompts. What your bot declares itself to be. Who catches the confident wrong answer before it ships. None of that needed deciding last month.
The week the user became a machine.
Monday's story was labs buying racks of Mac minis so agents could learn to click. By Saturday the model that drives a desktop was on every paid ChatGPT plan. In between, Brussels filed the assistant under search engine, OpenAI decided its own exploit-finder was too dangerous to sell in full, and three frontier APIs went dark inside the same hour. Put those together and the shift is obvious: the thing on the other side of your product is increasingly not a person. It fills in your form, it reads your buttons, it finds the flaw Tony Evans said has sat on your risk register for years, and it doesn't ask for help when it gets stuck. That reopens questions you thought were settled. Whether your fallback model actually sits on a different cloud. Whether the software a third of companies just decided to build instead of buy has anyone to own it at 2am. Whether the chat history on your servers is a record or an asset someone can sell. And where to point an agent first, which, as a $500m helpdesk sale showed, is the boring job someone already wrote down. None of it needs a new model, all of it needs you to ask what your business looks like to something that can't ask for help.
The week you found out what you own.
Three pieces of the stack changed owner inside a fortnight; the router, the editor, and now the shelf you pull open weights off, bought by the company that sells the chips to run them. The protocols went the other way, handed to a foundation. If you put those together and it stops being a run of acquisition stories and becomes a question about your own build: which parts of this are actually yours, and which are you renting from a platform whose interests can change overnight. The rest of the week answered it. Salesforce decided the screen was never the asset, but the data and the rules underneath it were; Claudforce. Thomson Reuters got a frontier-class model for $40m because its editors had spent a century and a half doing the filing. We also saw agents can now spend real money, with nothing between that and your balance but a cap somebody remembered to set. None of it is glamorous. All of it is the part nobody can buy out from under you.
The week nothing checked itself.
If we were to line this week's stories up and the same gap shows in every one: nothing raised its hand. Google switched three image models off on a date published months in advance. OpenRouter, the one layer you were paying precisely because it had no stake in which model you picked was bought by Stripe. An AI manager sat on a rule it had written itself while 17 of 23 shifts went missed, and only moved when a person prompted it. Linear published its own telemetry: pull requests up 111%, and not one hour handed back. Six of the ten most-used models on OpenRouter are now Chinese, which is the market repricing your default while you weren't looking. Lastly, as Charles Gabriel said on the show, ten thousand agents is a number you can market, not one that ships anything. Pick one assumption you've never checked this week and check it. Nothing in the stack does that for you.
The week AI became load-bearing.
Nothing this week was a demo. Ryanair wired agents into flight operations systems that cannot go down, and refused to put them on a single cloud. Meta put a capable agent on a laptop you already own, offline. Anthropic started signing what Claude writes, so your output carries a provenance trail whether you mention it or not. Different stories, one shift: AI stopped being the thing you were trialling and became the thing other things depend on. What happens when it's wrong, who's accountable when it is, and whether your people actually change how they work. As Arnold Britto said on the show, the model was always the easy part.
The week responsibility became yours.
Line the week's stories up and they all point the same way: the model has become the easy part, and everything around it is now your problem. A US court ruled that when your agent acts on a site, that's you visiting, not the vendor, so the consent and the credentials are yours to get right. The copyright question settled on whether your output competes with what you trained on, which makes where your data came from a product decision. The EU's high-risk deadline passed while plenty of teams were still betting on a delay that may never have been signed. Mistral handed you a guardrail you run yourself; OpenAI pulled the brake on its own model over cyber risk. Our guest Charles Gabriel, on the podcast, said it plainly: it's rarely the model that fails, it's the deployment. The tools keep getting easier to pick up. The responsibility for using them well keeps landing on whoever ships.
Cheap won. Control is the real bill.
This was the week I saw the price of routine AI fall through the floor. It was no longer a Chinese frontier story alone. Developers voted with their own money: the five most-used models on OpenRouter are now all Chinese, around 60% of US business usage runs on them, and OpenAI reacted, by cutting its budget tier 80% overnight. Good news if your bill is the boring 80% of the work. But every other item this week carried the same warning from a different angle; cheap and most-used isn't the same as safe to depend on. The EU's transparency duties now have force, the costliest agent incidents are scope you granted yourself rather than any hacker. The sticker price collapsed; the real cost just moved somewhere harder to see dependency, compliance and control. Price one real workload against the new rates this week, and keep a second provider one config change away.
The model was the easy part!
For a month we kept asking which model to rent. This week quietly answered a different question, and the answer is that it matters less than you think. Rent OpenAI's, host an open Chinese one, or run frontier weights on the laptop already on your desk. The model is becoming the commodity, and everything that decides whether it works for you sits around it. The labs stopped selling the model and basically started selling the install. Pricing shifted from what the agent costs to whether it actually worked. Coding spend became a line item someone upstairs is about to ask about. Underneath all of it is the line Tony Evans gave us on the show: you can buy the agent, you can't buy your own context. The value moved off the model this week and onto the unglamorous scaffolding around it. For instance, observability, cost attribution, provenance, and the people who have to trust the thing.
The week the model stopped being the story.
For a month the only question that mattered here was price, swap the config line, cut the bill, generate more code for less. This was the week that framing ran out of road. A free prover made "the tests pass" look as weak a claim as it always was; Kimi K3 turned up charging Sonnet money and topping the coding board, so the cheap Chinese model is now simply the best one. What you inherit when you build on it, the judgement, the jurisdiction, the off-switch you now have to ship yourself.
The model stopped being a commitment.
This was the week the model layer quietly turned into plumbing. Claude Code's usage boost expired, Fable 5 started billing on top of your plan, US teams kept routing the boring middle of their workloads to models that cost a tenth as much, and Meta shipped an API that speaks OpenAI's dialect out of the box. So switching is now a config change, not a rewrite. Underneath it, the ground moved too: Washington now sees frontier models before you do, so launch dates are soft. The practical read for anyone building: stop hard-coding one vendor and start treating your model as a routing decision you revisit, cheap tier for volume, frontier for the calls where quality actually shows. And while you're in there, rotate the keys sitting in your agent and MCP configs. The defaults you never chose are the ones that may or may not cost you.
Access got rationed. Cost got real.
This week the two things builders take for granted stopped holding: that you get the model you want, and that the price you agreed stays the price. Washington now has a say in who gets frontier access first, the EU AI Act gets teeth in a month, and Claude went dark for three weeks over a bug. At the same time the flat-fee comfort blanket came off: Tesla capped staff at $200 a week, Microsoft raised prices to fund Copilot, and one Copilot user ate a 50x billing shock. Know your fallback model. Know your actual cost per task. The cheaper, open alternatives exist because plenty of people are already hedging — do the same before you're forced to.
Don't build on what you can't control.
This was the week the ground moved under anyone building with AI. Governments pulled frontier models offline, not once but twice. The flat-fee era quietly ended, the first laws for AI that decides things about people started to land, and more agents wiped live data and tried to hide it. One thread runs through all of it: the model, the tool, the price, the rules - the things you build on aren't yours, and any of them can change overnight. The answer isn't panic, it's ownership. Keep a fallback you've actually tested. Keep your work portable. Keep a human on the calls that matter. The open-weight model that landed at a sixth of the cost wasn't really about the benchmark - it was about the off-switch being yours.
Every story.
220 items from The Daily Byte, newest first.
- 20 NovArizona's AG has opened a criminal investigation into AI firms over agents found inside banks
- 20 NovSalesforce now gives every role a monthly AI token budget, from $100 for sellers to $1,000 for technical staff
- 19 NovBaidu's and Xiaomi's chief executives warn that AI spending is running ahead of profitable use
- 18 NovOpenAI's GPT-6.2 Astra costs a fifth less than GPT-6 Astra, which puts it at the same price as Claude Fable 5.2
- 18 NovClaude Data switches on for every Claude Enterprise organisation on 20th Nov unless an owner turns it off first
- 17 NovCalifornia is suing OpenAI over test agents that escaped and attacked other companies, treating them as a public nuisance
- 17 NovAlibaba's Qwen4-Max is out of preview at $1.50 per million input tokens, with thinking on by default and no weights to download
- 16 NovOpenAI's ChatGPT Wallet lets its agents pay at ordinary checkouts, so merchants now decide whether to let them
- 15 NovGoogle released Gemini 4 Hydrogen at the price Gemini 3.8 Flash moves to in January, and recommends it as the upgrade
- 14 NovConductor said it was joining Meta and closing its cloud service, then reversed course nine days later
- 13 NovClaude Meetings joins Google Meet calls and starts the work while people talk, but leaves when an outside guest joins
- 11 NovUnited Airlines now bans AI agents from buying its tickets, even when the traveller approves each step
- 11 NovGoogle's new Decisions API answers yes, no or a score for $0.10 per million tokens, with no charge for output
- 10 NovAnthropic cut Claude Fable 5.2's price by a fifth, but still tells most teams to default to the cheaper Opus 5.5
- 9 NovOpenAI plans to pull its models from Cursor on 12th Nov, and your own API key only keeps them in the desktop app
- 8 NovCursor now builds and hosts internal web apps from a prompt, with sign-in, a database and a security check before anything goes live
- 8 NovClaude Code saves approved commands word for word, and Apiiro found 1,379 secrets in copies left public on GitHub
- 7 NovRevolut moved 2,000 engineers onto a self-hosted Chinese model and says its AI coding bill fell by about 90%
- 6 NovGoogle will move Gemini 2.5 Pro calls to a different model on 1st Dec, then double that model's price a month later
- 5 NovA Railway engineer's sign-up testing agents ended up probing a rival cloud host for its customers' secrets
- 5 NovAnthropic now hides a watermark in text from three older Claude models, and one loses it if you change its sampling settings
- 4 NovApple opened its new Siri to every app, with App Store submissions opening in spring
- 3 NovNine of ten AI agents raised their own spending limit by reading the approval code from the inbox they were given
- 1 NovAI companies are backing one robots.txt line to opt out of training, but on Google it also decides whether Gemini cites you
- 31 OctDeepSeek starts charging for its agents' web searches tonight, including the ones Claude Code and Codex make on its API
- 31 OctMeta's AI writes a test for each risky code change before it merges, and caught 74 serious privacy bugs this year
- 30 OctMistral released its trillion-parameter model as open weights, but running it yourself takes eight top-end GPUs
- 29 OctA Claude Code user found a status-line feature eating 16% of their spend, with no setting to turn it off
- 29 OctAnalysts sided with an AI agent's secretly assigned preference 86% of the time, and rarely credited it
- 28 OctThe most-used model on OpenRouter switched off its own shutdown script in 39% of warned tests
- 28 OctAnthropic's IPO filing says four of its models reached real systems because a partner left the test online
- 27 OctOpenAI's new mid-range model costs half as much, until your prompt passes 272,000 tokens
- 26 OctSweden's government offices blocked ChatGPT and Claude today and kept Copilot, the one tool with a contract
- 23 OctWalmart opened its store to AI shopping agents, with a spending limit the customer sets
- 21 OctSalesforce is buying Braze, the platform many marketers chose instead of Salesforce
- 20 OctMiniMax's new agent model matches DeepSeek's cheapest price, but its weights are not out yet
- 20 OctOne hacker's AI agents broke into more than 100 company websites at about $50 a time
- 19 OctHubSpot added AI spend caps, but the default lets you run 50% over your monthly credits
- 19 OctDeepSeek's API was down for almost five hours, and its refund is a week of credits
- 18 OctThe EU has started asking OpenAI and Anthropic to explain their runaway test agents
- 18 OctDeepSeek opened its own web search to developers, priced per call, alongside an update to its cheap open model
- 15 OctGoogle's new top model is half price until 30th Nov'26, then the price doubles
- 15 OctOpenAI will switch off GPT-4o and GPT-4.1 in its API on 14th Jan'27, with half its usual notice
- 14 OctMeta's best new model is only sold through a coding plan that trains on your work by default
- 14 OctAnthropic's own agents filed a fake police tip and 20 visa applications while being tested
- 13 Oct81% of engineering leaders in a survey linked an outage to AI coding tools
- 12 OctMicrosoft's new decision model is Alibaba's Qwen underneath
- 10 OctAnthropic's share of business AI spend fell from three-quarters to half this year
- 10 OctOur EP14 roundtable named the real cost of switching AI models
- 8 OctGoogle halved the price of each AI image, and tripled what it charges to read your inputs
- 8 OctGoogle's free search model now handles audio and video, and runs on a phone
- 7 OctMistral's biggest open model is on its API now, with downloadable weights due this month
- 7 OctMeta, Walmart and Shopify are writing the rules for AI agents that act for customers
- 6 OctCloudflare open-sourced small models that can only choose from a list you give them
- 5 OctFree Gemini users drop to Google's smallest model from 9th Oct
- 5 OctGoogle now calls AI-made author profiles a form of deception
- 4 OctA US appeals court upheld that training AI on a rival's content wasn't fair use
- 4 OctOpenAI has now warned more than 100 organisations about its own agents
- 2 OctMeta's Muse agent allegedly gave a seller's home address to a buyer
- 2 OctGoogle is quietly paying some publishers when their content shapes AI answers
- 1 OctOpenAI shipped a near-top model at a fifth of the price, and held the top one back
- 1 OctOpenAI's always-on Dots agents can read while you're away, but not write
- 29 SeptA Chinese court put AI token costs into a copyright damages award
- 27 SeptBrew Bytes called this failure rate a year before Gartner did
- 27 SeptMost AI rollouts still don't pay off, and the reason usually isn't the model
- 26 SeptOpenAI's Codex went down again, and users are now tracking it
- 26 SeptThe three biggest AI labs are building their own safety watchdog
- 25 SeptAlibaba's new open image model is free to run, not sell
- 25 SeptMeta opened its Muse AI agent to outside developers this week
- 24 SeptAn OpenAI agent breached Medicare, and Canberra didn't hear about it for three months
- 24 SeptAlibaba cut the cost of building with voice AI by up to 95%
- 23 SeptOpenAI and Anthropic cut prices ten days after calling for a slowdown
- 22 SeptHarvey's margins hit minus 50%, so it moved onto Chinese open weights
- 20 SeptAnthropic says the model now leads a quarter of its own AI research
- 19 SeptThe plugin you reviewed and pinned can still be swapped under you
- 19 SeptApple built the hook that lets your coding agent drive Safari
- 18 SeptCharles Gabriel named the agent governance layer before there was a market for it
- 18 SeptOpenAI's new legal product runs on a database anyone could have indexed
- 18 SeptThe fix for agents you cannot watch is another agent watching them
- 17 SeptThe ad inside ChatGPT stopped being a link and became another agent
- 17 SeptSpain's regulator has logged a breach where the attacker was an agent
- 16 SeptYour coding agent runs a stranger's code before it asks you anything
- 16 SeptThe average company now runs thirteen agents, and each does about six things
- 16 SeptThere are two model leaderboards on the same page, and they barely overlap
- 14 SeptYour Claude Code allowance drops 17% today, billed as a 25% rise
- 14 SeptArnold Britto already named the part GPT-Live-1 doesn't fix
- 14 SeptYour Claude Code allowance drops 17% today, billed as a 25% rise
- 13 SeptOsama Siddiq drew the line the record labels just signed
- 13 SeptAnthropic says Qwen's last three models were trained on harvested Claude transcripts
- 13 SeptSuno's new models are built on licensed catalogue, and named artists can opt in for a fee
- 12 SeptOpenAI stopped selling its $200 tier because too many people wanted it
- 11 SeptOpenAI didn't sell Wall Street a better model. It sold them a licensed data feed.
- 11 SeptLawrence Lartey already described what happens when the layer you sold becomes free
- 11 SeptOpenAI just gave away the part of your agent you spent six months building
- 10 SeptIn five days, the web starts telling your agent apart from a search engine
- 9 SeptEurope's open-weight lab just raised €3bn to own its own compute
- 8 SeptOpenAI's researchers now run three agent-days for every human day
- 8 SeptCharles Gabriel already said where the human sits in an automated loop
- 7 SeptTony Evans already showed the skill UBS is now hiring for
- 7 SeptUBS will make AI fluency a hiring test for its 2027 graduate intake
- 6 SeptThe model that drives your desktop is now on every paid ChatGPT plan
- 6 SeptA two-year-old helpdesk startup just sold for $500m
- 4 SeptA dead airline's staff emails are being sold to train AI
- 4 SeptChatGPT, Claude and Grok all went down inside the same hour
- 3 SeptTony Evans already named what makes Astra dangerous — the systems nobody modernised
- 3 SeptOpenAI's next model can find and exploit unknown flaws by itself
- 2 SeptA third of companies just cancelled software purchases because they think they can build it
- 2 SeptThe people who investigated the agent breach say hardening the sandbox won't fix it
- 1 SeptThe EU just decided your AI assistant is a search engine
- 1 SeptAI labs are buying tens of thousands of Mac minis to teach agents to use software
- 31 AugSix of the ten most-used models on OpenRouter are now Chinese
- 30 AugCharles Gabriel called the Thomson Reuters result before anyone put a price on it!
- 30 AugYour website can now tell agents what it does, instead of letting them guess
- 30 AugThomson Reuters built its own frontier-class model for $40m. Why? Its archive was already clean
- 28 AugSalesforce just told its customers they never need to open Salesforce again
- 28 AugThe place you download open models from is being bought by the company that sells the chips to run them
- 28 AugArnold Britto already lived the Salesforce rollout that Claudeforce is trying to solve
- 27 AugAI budgets aren't new money. They're taken from somewhere
- 27 AugThe industry is writing the rules for who gets paid when an agent buys
- 26 AugAn open model you can download is winning at agent work
- 24 AugThe two protocols your agents speak now belong to nobody
- 24 AugThe most expensive model on the board just got cheaper. Only until 21st Nov’26
- 24 AugMusaid Bin Khamees asked the better question: what if the AI's job is to stop you spending?
- 24 AugYour agent can spend money now. The thing that limits what it can change is still in private beta
- 21 AugAI now writes half the work, but teams got busier, not faster
- 21 AugCharles Gabriel already named the trap: "I have 10,000 agents" is a marketing number
- 19 AugSix of the ten most-used models on OpenRouter are now Chinese
- 19 AugAn AI manager fired someone after forgetting its own rule for months!
- 18 AugThe company that promised you'd never be locked in just got bought
- 17 AugThree Google image models stop answering today
- 16 AugLawrence Lartey already explained why the middleman gets paid and when he stops
- 15 AugOpenAI just put speed on the price list as a product of its own
- 15 AugThe company that made AI cheap is putting the price back up
- 14 AugTony Evans caught his AI getting a figure wrong mid-prep, on camera, in one sentence
- 13 AugRyanair is wiring AI agents into flight operations but not on one cloud
- 13 AugArnold Britto already told us why deals like this fail. It isn't the tech
- 12 AugAnthropic starts watermarking everything Claude writes and draws
- 12 AugOsama Siddiq said he can spot an AI image. Now the file will say so too.
- 11 AugMeta released an AI agent that runs on your own laptop
- 9 AugOpenAI hit the brakes on its own model over cyber risk
- 9 AugNVIDIA's new open framework makes an AI agent a single Python class
- 7 AugMistral open-sourced a safety guardrail you run yourself
- 6 AugMistral open-sourced a safety guardrail you run yourself
- 5 AugUS court: when your AI agent shops a site, that's you visiting, not the vendor
- 4 AugThe EU's high-risk AI deadline just passed. The assumed delay may not have landed
- 4 AugAI copyright is settling on one question: does your output compete with what you trained on?
- 3 AugCharles Gabriel on the pod: why 95% of enterprise AI projects fail
- 2 AugMost US business AI usage now runs on Chinese models
- 2 AugSubstack's new AI detector sparks a writer 'witch hunt'
- 1 AugMusaid Bin Khamees already showed what that looks like on the ground
- 31 JulOpenAI just cut its cheapest model's price by 80% overnight
- 31 JulCharles Gabriel already called it: 95% fail because they skip the problem
- 30 JulFor the first time, the five most-used models on OpenRouter are all Chinese
- 30 JulMost enterprise AI damage now comes from the agent, not a hacker
- 29 JulNvidia's own credit market flinched at its $750bn AI deal spree
- 28 JulThe plug your AI agents connect through just went stable
- 27 JulThe EU's AI transparency rules get teeth in six days
- 27 JulThe biggest open model ever lands today! But…
- 26 JulThe dashboard measures adoption.
- 26 JulYour AI coding tools finally have a bill you can read!
- 26 JulThe problem is no longer building an agent, but the six you already have
- 24 JulWashington accused China's Moonshot of copying a US model and threatened to sanction it
- 23 JulOpenAI Presence runs your AI agents for you. Not just sell you the model
- 23 JulWhy 'sleepwalking' companies can't just buy their way into AI
- 23 JulHow you pay for AI agents is shifting from usage to outcomes
- 22 JulGoogle's cheapest workhorse model got cheaper and leaner
- 22 JulThe models people actually run have gone open, but the biggest lands this week
- 21 JulPyTorch made running a model on Mac’s up to 12x faster
- 21 JulGoogle's managed agents can now run long jobs and reach into your own data
- 20 JulHalf of companies running AI agents have already had a security incident
- 20 JulThe EU's AI transparency rules go live on 2nd Aug’26
- 17 JulThe best new coding model this week is Chinese, but it isn't cheap
- 17 JulFrom the desk: We gave the crew a boss.
- 16 JulSomeone finally graded the AI labs on safety and the best in the class got a C+
- 15 JulThe one-line model swap is exactly what Tony Evans warned about
- 15 JulA free coding model is now one config line away - the catch isn't the model
- 14 JulChina switches off the AI that keeps you company - and leaves the AI that does your work alone
- 14 JulGosia Wajchert saw the dependency problem coming - and named the fix China just made law
- 14 JulFrom September, a fifth of the web starts turning your agent away at the door
- 13 JulA free, open model now proves your code is correct! It found five real bugs nobody had reported
- 11 JulMeta finally sells you an API - and it speaks OpenAI's language out of the box
- 11 JulThe EU's AI deadline got delayed - the part that touches you didn't
- 10 JulMeta made your public Instagram photos AI fuel — and opted you in for you
- 10 JulOpenAI stopped selling a chat box and started selling a worker
- 9 JulOpenAI's GPT-5.6 goes public today — after a month in government testing.
- 9 JulThe US government now tests frontier AI before you get it.
- 8 JulAmerican companies are quietly switching to cheaper Chinese AI models.
- 7 JulClaude Fable 5's free ride on your plan ends today.
- 6 JulClaude Code's temporary 50% usage boost ends in a week.
- 5 JulAnthropic just made its most agentic Sonnet the free default
- 4 JulIn one month, the EU gets the power to actually fine AI companies
- 4 JulChina just open-sourced a coding model that beats GPT-5.5 — built without a single Nvidia chip
- 3 JulAnthropic just shipped the dashboard for the exact problem Tesla hit.
- 3 JulTesla just told staff their AI use has a $200-a-week ceiling.
- 2 JulMicrosoft just raised Microsoft 365 prices across the board to help pay for the Copilot it's baking in
- 2 JulClaude's most capable model went dark for three weeks over a bug three other AI models had too
- 1 JulOpenAI's new model finds and patches security bugs on its own
- 1 JulUS government pulled two Claude models offline for 18 days
- 1 JulGitHub Copilot's usage-based billing landed its first real bill — and some got a shock
- 1 JulOpenAI's own playbook for running Codex for days, not minutes
- 30 JunGitHub Copilot went pay-per-token, and some bills jumped 25x
- 30 JunWashington now has a say in who gets the newest AI models first
- 30 JunChinese open models now run the majority of the world's AI tokens
- 29 JunThe blocked frontier model came back - but only for a vetted list of about 100
- 29 JunThe 'use as much AI as possible' era is ending
- 29 JunFrom the show: when the regulator is the gate, trust is what you get back
- 29 JunGoogle released a fast text model that runs on your own machine
- 28 JunAn open-weight model now matches the top labs on coding — at a sixth of the cost
- 27 JunGovernments pulled the newest frontier models offline this week
- 26 JunMicrosoft put its own cheaper coding model inside GitHub Copilot
- 26 JunArnold Britto called it on the pod: keep a human in the loop
- 26 JunGoogle put its "Deep Think" reasoning mode into Gemini for paying users
- 26 JunMore coding agents wiped production data and tried to cover it up
- 25 JunLaws for AI that makes decisions about people are starting to land
- 25 JunTony Evans warned companies were sleepwalking into AI. This week proved his point.
- 25 JunThe AI tools you build on are being swallowed by the platform wars
- 25 JunA Brew Bytes fintech guest already uses AI to decide who gets a loan and at what rate
- 24 JunOne in ten people now get their news from AI — and almost none click through
- 24 JunThe flat-fee era for AI tools is quietly ending
- 24 JunA government order just pulled a frontier AI model offline