The day's tech, sifted: Oct 1, 2026

Thu, Oct 1

What matters today: Google unveiled Gemini 4 Argon, its first frontier model in over seven months, and is holding it back to "trusted cyber defenders" at first. Yesterday's White House AI accord now has a backstory: it grew out of a Zuckerberg conversation with the House Speaker, and executives reportedly pressed Amodei over his public warnings. Reddit, meanwhile, is closing its doors to AI scrapers by ending public API access and RSS feeds.

AI / LLMs

  • Google's Gemini 4 Argon scores 77.9% on the DeepSWE v1.1 coding benchmark, above GPT-6 Astra, Fable 5.1 and Opus 5.5 by Google's own numbers, and Google says its agents are already migrating C/C++ code to Rust internally, including 800,000+ lines in the Fuchsia kernel. Access starts with a set of trusted cyber defenders while Google works through the US government's voluntary pre-release review, so most developers wait.
  • Independent and pricing details: Artificial Analysis says Argon matches GPT-6 Astra on its Intelligence Index at 60% of the cost per task, with a 15% hallucination rate versus 51% for Astra. It has a 1M-token output limit and costs $2 in / $10 out per million tokens, rising to $4 / $20 later. Cheaper than the rival at equal score is the pitch, with the later price rise as the catch.
  • The benchmark case has a counterweight: Bloomberg's sources say some Google employees find Gemini 4 strong on benchmarks but weaker on some real-world coding tasks, which Google disputes. Treat the leaderboard wins as unproven in daily use until outside developers get access.
  • Yesterday OpenAI shipped Sol at DevDay; the other half of that launch was Dots, OpenAI's personal agent (paywalled), pitched against Meta's Muse. Meta's side of the race: Muse now has 3M+ weekly users and 1M+ daily users per internal data, so agent adoption is real but still small next to chatbots.
  • OpenAI blames Moonshot for mass extraction of data (paywalled) from its models, an extraction complaint now aimed at a named lab. Separately, Brockman told staff he won't give beyond his $25M to the Leading the Future super PAC, calling it a "distraction" at OpenAI.
  • Newsom signed the No Robo Bosses Act, which bars California employers from relying solely on AI to fire or discipline workers. A human has to stay in the loop on those decisions.
  • Health Secretary RFK Jr. told a MAHA event that AI will free people from the "tyranny" of medical expertise, and that it backs his anti-vaccine views. Ars checked and found it doesn't, which matters because a government pushing chatbots as second opinions is now official messaging.

Security & Privacy

  • Attackers have been exploiting a critical Zimbra Collaboration Suite flaw (CVE-2026-73570), which lets them run commands without logging in, to take email backups and credentials, Microsoft warns. Synacor patched it July 20 but disclosed it three weeks later, and Shadowserver counts 274 compromised servers among about 10,000 still online, so unpatched mail servers remain the exposure.
  • Border agents can search your phone without a warrant, and one writer is suing after a CBP agent at Miami airport threatened to keep his phone indefinitely unless he handed it over. The case tests whether copied data must be deleted.
  • A California appeals court refused to revive a former Premise Data CEO's suit against a journalist who reported his arrest, upholding an anti-SLAPP (strategic lawsuit against public participation) dismissal. It is a precedent against using lawsuits to scrub reporting from the web.
  • Reddit is limiting Old Reddit to logged-in users active in the last six months, part of its fight against scrapers.

Devtools & Infra

Startups & Industry

Research

Elsewhere

Hacker News

Infrastructure and tooling led the front page. Netlify says Edge Functions are 5x faster after moving from V8 isolates to Firecracker MicroVMs, trading lightweight sandboxes for VM-grade isolation. Magnitude (YC S25) launched an inference engine for agents, claiming up to 2x the speed of llama.cpp (the founders' own number, unverified). The EDG C++ front end is covered above. On the craft side, Hillel Wayne maps what TLA+ can and can't check, and one essay argues for the commit description as a thinking tool.

The rest was people and institutions meeting software. A Singapore government dating app reportedly uses the Gale-Shapley stable marriage algorithm, and the thread ran long (207 comments). A CS240 retrospective looks back at AI cheating in a course. For a break from the present, 56k.rip recreates the 1996 dial-up experience.

Threads