AI news, with practical context.
Most AI coverage is press-release laundry: a new model drops, sites race to repackage the announcement, nothing changes for the person actually building. We do the opposite. We publish when something in AI — a release, a pricing change, an agent platform, a policy shift — is worth doing something different about. Every post cites its sources, separates fact from spin, and ends with a clear point of view on who should care and what to change in their stack.
What we cover
Four lanes. Everything else gets skipped on purpose. If a story doesn’t change what a practitioner should build or buy, it doesn’t make the page.
Model releases
Frontier and open-weights models from OpenAI, Anthropic, Google, Meta, Mistral and Qwen — but only when capability, price or latency actually shifts where you should run inference.
In /ai/models/Agent platforms
New launches and feature drops on Lindy, Gumloop, Relevance AI, Zapier Agents, n8n and Make. We track what changes for solo founders and small teams, not enterprise procurement.
In /ai/agents/Pricing & API changes
Per-token pricing moves, rate-limit shifts, plan restructures and quiet contract changes that affect your monthly bill. The category vendors hope you don’t notice until renewal.
Impact analysisSecurity & policy
Data-handling policy changes, training-data terms, prompt-injection disclosures, EU AI Act milestones and other shifts that quietly change what you can and can’t deploy.
Risk & complianceThe sources we track
Six recurring beats we monitor every week. First news posts go live shortly — in the meantime, every card links to the umbrella AI hub where the deeper coverage already lives.
OpenAI changelog watch
Model releases, API behaviour changes, pricing tweaks and quota shifts on the OpenAI platform. We track this source weekly. First posts going live shortly.
Browse AI coverageAnthropic release notes
Claude model checkpoints, Claude Code feature drops, computer-use changes and policy updates from Anthropic. We track this source weekly. First posts going live shortly.
Browse AI coverageGoogle AI updates
Gemini model releases, Vertex pricing, Workspace AI features and DeepMind research worth acting on. We track this source weekly. First posts going live shortly.
Browse AI coverageMeta + open-source models
Llama releases, Qwen, Mistral, DeepSeek and the broader open-weights ecosystem. Where the local-inference and self-hosted story changes. First posts going live shortly.
Browse AI coveragen8n + Make + Zapier feature drops
Native AI nodes, new triggers, agent-builder features and integration updates across the three workflow platforms most readers already run on. First posts going live shortly.
Browse AI coveragePricing & API change tracker
The quiet per-token price moves, plan restructures, rate-limit shifts and grandfathering changes that only show up on your bill three months later. First posts going live shortly.
Browse AI coverageHow we cover news differently
A short explanation of the editorial rules behind every post on this page, so you know what to expect before you bookmark us.
We don’t blog every press release
The AI news cycle is roughly five real stories a month wrapped in five hundred filler ones. Vendor announcements, UI refreshes, “partnership” releases, leaderboard movements that don’t reflect real-world quality — most of it changes nothing about what you build on Monday. Our editorial bar is a single test: would a thoughtful practitioner do something different because of this? Most stories fail that test. We skip them. That filter is the entire reason this hub exists, and it’s what separates us from sites paid per page-view.
We publish when something changes what you should do
When we do publish, it’s because the story has a practical edge for an entrepreneur, developer or creator. A 30% price drop on a frontier API that flips the build-vs-buy math. A new agent feature that removes a manual step from a common workflow. A policy change that retroactively affects what you can run in the EU. A model that beats the previous best at a specific job — coding, long-context, vision — by enough that it changes the recommendation in our best picks library or the comparison logic in our comparisons hub. That’s the bar.
Sources cited, screenshots taken, links live
Every news post lists its sources at the bottom, with direct links to official changelogs, blog posts, pricing pages, GitHub release notes or regulator filings. When we cite a price, we screenshot the pricing page on the day we publish — vendors revise pages quietly more often than they admit, and we want a paper trail. When we cite a benchmark, we link the benchmark and note the version. When we quote a vendor, we link the original. The aim is that you could fact-check any sentence in any of our news posts in under sixty seconds. If we ever get something wrong, we say so at the top of the page with a dated correction, not a stealth edit.
Every post has a clear “our take”
The structure we follow for every news piece is consistent on purpose. What happened — the news in two paragraphs, dated and sourced. Why it matters — the practical reason it’s not noise. Who should care — solo founders, small teams, technical builders, content creators, or some combination. Practical impact — what changes in your stack, your bill, or your roadmap if you accept the news. Our take — what we’d actually do this week, named opinions, not hedged. Sources — links and timestamps. You can read the “what happened” if that’s all you need, or scroll to “our take” if you trust us by now and just want the recommendation. The rest of the post is for the readers who need to verify the chain of reasoning before changing their own systems — and they’re the readers we care about most.
One more thing we don’t do
We don’t run sponsored news posts. Vendors can advertise on the site, but no money buys coverage in this hub. If a tool we cover is also one we recommend in our agents or automation hubs, we say so in the post. If a news item helps a vendor we don’t recommend, we still write it honestly. The whole reason the hub exists is that the rest of the AI press doesn’t draw that line clearly enough.
Get our weekly AI-news roundup
One short email a week. Only the AI news that actually changes what to build, with our take and the sources. No filler, no roundup-of-roundups, no “10 prompts that will change your life.” When the email goes live we’ll add a signup form here — until then, the best picks library is the closest thing.
Explore best picksEmail signup launches with the first wave of news posts.
Honest answers about the editorial bar
Four questions readers ask before they decide to come back regularly. Straight answers.
Irregularly, by design. We publish when something changes what entrepreneurs, developers or creators should actually do. In a quiet week that’s nothing. In a busy week — a frontier model release, a major pricing reshuffle, a policy shift — that’s three or four posts in a few days. We’d rather skip a launch than write a piece that adds noise. There are already enough sites covering every Twitter announcement, and we’re not trying to compete with them on speed. If you want hourly news, follow the changelogs directly. If you want filtered, sourced takes on the news that matters, this page is the one.
Official channels first: company changelogs, vendor blogs, pricing pages, status dashboards, GitHub release notes, regulator publications. We then cross-check with the practitioners we already know building on these platforms — independent operators on Lindy, Gumloop, n8n, Make and Zapier who’ll tell us when a feature is real and when it’s vapour. Every news post lists its sources at the bottom with direct links. If we cite a benchmark, we link the benchmark and the version. If we cite a price change, we screenshot the pricing page on the day we publish, because vendors revise pages quietly more often than they admit.
Because most announcements are noise from a buyer’s perspective. New model checkpoints with no measurable benchmark movement, UI refreshes, rebrands, “strategic partnership” press releases, integration announcements between tools you don’t use — they fill feeds and change nothing about how you should build next week. Our editorial bar is a single question: would a practitioner do something different tomorrow because of this story? If the answer is no, we skip it. That filter throws out maybe 80% of what trade press covers and frees us to write properly about the 20% that matters. That’s the whole product.
Probably, eventually, if it ships something that changes how people should build. We’re deliberately not committing to coverage of every vendor — the news hub is curatorial, not comprehensive, and we’d rather underpromise than greenwash our publication schedule. If you want to flag a release we’ve missed, send the official source link and a one-line case for why it changes a workflow. We read every one we receive; we publish maybe one in five. Volume is not the metric we’re optimising for, and that’s the deal. Honesty about the bar is part of what makes the hub trustworthy when something does land.
