Daily Dispatch · Artificial Intelligence & TechnologyNoon · Eastern · No. 3

The Tech Roundup

Every morning · One email

We read 40 tech newsletters at dawn so you can read one at noon.

The Tech Roundup folds the entire morning of AI and tech news into a single daily edition — the numbers, the contradictions reconciled, and a take on why it actually matters. No doomscrolling. No “big if true.” No 47 open tabs.

40+Newsletters read
570+Primary sources
1Edition / day
0Hype, ever
Why it’s different
01

Synthesized, not stacked

Twelve takes on the same launch become one clear story — the marketing stripped out, the conflicting numbers reconciled, the duplicate links collapsed to the original source.

02

The “so what,” every time

Every story ends with why it matters for an actual human — not just what shipped, but what it does to your job, your tools, and your week. That’s the whole point.

03

The thing nobody surfaced

An Under the Radarslot for the consequential story that didn’t trend — scored by a panel of rival AI models, so importance wins over a popularity contest.

A real edition

Latest Edition

Saturday · June 6, 2026

Top Stories — 15 · Under the Radar — 3 · Hype — 0

AnthropicCovered by 6 sources

Claude now writes 80% of Anthropic's code — and Anthropic says that's the scary part

Anthropic published research reporting that Claude authored over 80% of the production code merged at the company in May 2026, with the average engineer shipping 8x more code per day than in 2024 — and Claude's success rate on internal engineering tasks reportedly jumped from 26% to 76% in six months. The company frames this as an early sign of recursive self-improvement: AI helping build the next AI, fast enough that it could arrive before institutions are ready. The pitch comes with a pledge to slow or pause frontier development — but only if rival labs agree to do the same.

Why it matters

The convenient thing about a safety warning that doubles as a flex about your own product is that it can't lose: either Claude is dangerously capable, or it's an impressive marketing line, and Anthropic gets credit either way. Worth noting the alarming numbers are all Anthropic's own and self-reported, and the proposed pause hinges on competitors who have shown zero interest in slowing down — which makes it less a brake and more a press release.

Source: Anthropic

Google DeepMindCovered by 6 sources

Google's Gemma 4 12B runs text, vision, and audio on a 16GB laptop — no cloud required

Google DeepMind released Gemma 4 12B, an Apache 2.0-licensed multimodal model that handles text, vision, and audio and runs locally on a standard 16GB laptop (about 8GB with quantization), with a 256K context window and tool calling. The trick is an encoder-free design that folds vision and audio straight into the LLM backbone, which Google says makes it nearly as capable as the bigger Gemma 4 26B MoE — and it's the first Gemma of this size to do native audio. Via Google AI Edge, it analyzes data and generates scripts on-device, so your files never leave the machine.

Why it matters

"Open weights you can actually run on the laptop you already own" is the part that matters here — no API bill, no upload, no someone-else's-server holding your documents. For anyone who balked at sending personal or work data to the cloud just to use AI, the privacy math finally tilts the other way: the model comes to your data instead of the reverse.

Source: Google DeepMind

OpenAICovered by 6 sources

OpenAI's Codex hits 5M weekly users — and non-coders are the fastest-growing crowd

Codex crossed 5 million weekly users, and OpenAI is rebuilding it from a developer tool into a general-purpose work platform: 110 skills across 62 apps, six role-specific plug-ins (sales, analytics, creative, product design, equity investing, investment banking), plus a new Sites feature that spins prompts into hosted, live-data web apps via partners like Figma and Wix. The tell is in the usage split — non-developers are now about 20% of users and growing 3x faster than the coders who got there first.

Why it matters

OpenAI started by automating the engineers; the growth curve says the spreadsheet-and-slides crowd is next, not safely behind them. When the fastest-growing users of a coding tool can't code, the company has quietly stopped building a dev product and started building a replacement for the desk job.

Source: OpenAI

◆ Under the Radar1 source · panel-picked

Researchers chained a Zapier sandbox escape all the way to NPM publish tokens

Token Security detailed an attack chain dubbed "Zapocalypse" that started by breaking out of Zapier's Python sandbox to run os.system, then pulled STS tokens off a Lambda's heap, and ended by lifting high-privilege NPM publish tokens. Strung together, that reportedly opened the door to account takeover and access to private repositories. To be clear, this is the researchers' write-up of what the chain *could* have done — framed as a hijack that was caught, not one that hit users.

Why it matters

NPM publish tokens are the keys to the software a huge chunk of the internet quietly installs every day, so an escape that reaches them isn't just one company's problem — it's a supply-chain problem. The unglamorous lesson: every "run user code safely" sandbox is one clever hop away from the credentials it was supposed to protect.

Source: token.security

Read the full edition (18 stories) →

Start tomorrow with the whole picture.

One email, noon Eastern. The morning’s AI and tech news, already read and made sense of.