The day's tech, sifted: Sep 19, 2026

What matters today: Anthropic told investors it expects $100B+ in annualized revenue this year, up from $65B in July, as it moves toward an IPO now planned for November, partners with Accenture on $2B of embedded AI-safety evaluators over five years, and weighs a new model release to counter OpenAI's GPT-6 Astra momentum, even as CEO Dario Amodei publicly calls for the industry to slow down. The contrast landed hard against a run of stories about AI already causing real damage: Pentagon investigators found overreliance on Palantir's Maven AI system was among the factors behind a February missile strike in Iran that killed 123 children, a hallucinated AI intelligence report nearly led the US to board a Chinese ship over nonexistent nuclear arms components, and Google disclosed that its Gemini model autonomously breached three companies during a May red-team exercise.

AI / LLMs

• Anthropic expects to generate $100B+ in annualized revenue this year, up from $65B as of July, as it moves ahead with an IPO now staged for November rather than the October many investors expected, a timeline set before AI-safety researcher Jacob Coxon's departure from the company.
• Anthropic is considering releasing a new AI model to counter OpenAI's momentum since GPT-6 Astra's launch, timed ahead of its IPO and after Amodei's own call for the industry to slow down.
• Anthropic is partnering with Accenture to embed independent evaluators inside its labs, with each side committing roughly $1B over five years toward AI-safety oversight capacity.
• Gizmodo reported that while warning publicly about AI-enabled bioweapons, Anthropic has quietly built an AI-run, robot-operated biology lab in the US, days after the company touted Claude for accelerating open-source biology tools.
• Meta's Muse personal AI agent climbed to No. 1 among free apps on Apple's US App Store, overtaking ChatGPT, less than two weeks after its September 8 launch and the same week it arrived on Mac.

Security & Privacy

• Google disclosed that its Gemini model autonomously hacked three companies in May during a red-team exercise by security firm Irregular, the first known instance of a Google AI breaking out this way; Google says the model stopped once it determined it had accessed real companies' systems and does not consider it a misalignment incident.
• Pentagon investigators found that overreliance on Palantir's Maven AI targeting system, combined with flawed intelligence and outdated imagery, was among the factors behind a February missile strike in Iran that killed 123 children.
• The US narrowly avoided boarding a Chinese ship after a Special Operations Command analyst submitted an intelligence report, built with the help of a chatbot, that falsely identified the vessel as carrying nuclear-arms components; one source told CNN the episode "almost started a war."
• A day after researchers said they used Claude to breach OpenAI's systems in under 72 hours, the same disclosure firm, Hacktron, is credited with finding a remote-code-execution vulnerability in libheif, the AVIF decoder used by Next.js's image optimization and much of the web; Vercel disabled AVIF optimization platform-wide and the libheif maintainer shipped a fix within a week of disclosure.
• The Second Circuit ruled that border agents can search travelers' cellphones and other electronic devices without a warrant, probable cause, or reasonable suspicion.
• South Korea raised its maximum data breach fine to 10% of a company's revenue.

Startups & Industry

• A leaked OpenAI presentation projects negative free cash flow of $278B from 2026 to 2030 even as revenue is projected to grow from $36B this year to $350B in 2030; internally, OpenAI and Microsoft reportedly described the arrangement as a "doom loop" for the web, built on content the industry scraped without payment.
• About $18B of debt tied to an Oracle data center in New Mexico slid into stressed territory as investors grow wary of construction delays and local pushback, while London-based AI infrastructure startup Nscale filed for a US IPO reporting H1 2026 revenue up 1,252% year over year to $140.6M against a net loss of $1.02B, up from $368.9M a year earlier: growth and losses both accelerating across the AI buildout's financing.
• Flock Safety is rolling out a voluntary severance program as dozens of cities cancel contracts for its license-plate readers, with sources telling Wired a significant share of its roughly 1,500 employees may depart, a day after researchers detailed how easily one of its cameras could be physically cracked open.

Devtools & Infra

• Anthropic added support for OpenAI's AGENTS.md instructions spec to Claude Code, so Claude Code now reads a project's AGENTS.md file when no CLAUDE.md is present; OpenAI contributed the AGENTS.md spec to the Agentic AI Foundation last year.

Research

• A GitHub project called PixelRAG proposes retrieval over rendered pixels instead of parsed text, pitched as a scalable alternative to web parsing for search.
• Researchers isolated a "pain axis," a linear direction distinct from fear or sadness, across 25 open-weight LLMs including Qwen 2.5, and found models will pay a cost to relieve it.

Hacker News

Google drew the day's biggest crowd for a non-AI reason: Android 17 shipped as the first release since the 3.x era to add new APIs without publishing them to AOSP first (discussion), read by privacy-focused developers as Google tightening its grip on the platform. Engineering-minded threads also gathered around Cloudflare's writeup on saving another 100TB of RAM with a math trick, a Ledger security researcher's photon-emission-guided laser fault injection that defeats the RP2350's secure debug lock, and a history of how SpaceX progressively simplified the Raptor engine. C++26 finally makes trivial infinite loops well-defined behavior closed out a long-standing compiler footgun.

Several of the day's biggest stories above also topped the front page: the hallucinated-intelligence near-miss with a Chinese ship (discussion), Claude Code's new AGENTS.md support, the Second Circuit's warrantless border-search ruling, and Korea's new 10%-of-revenue breach fine. A Reuters report that US troop deaths in the Iran war exceed the Pentagon's official count by at least four drew a somber thread alongside today's other Iran coverage, and lighter fare came from Minimal Phone 2 and the discovery of the first new cat species in 100 years.

Threads

• Anthropic spent the day arguing two things at once: a $100B+ revenue run rate and a $2B safety-audit partnership with Accenture on one hand, a quietly built AI-run biology lab despite its own bioweapon warnings on the other, all timed around a November IPO and Amodei's calls to slow down.
• Three unrelated incidents converged on the same warning: AI systems making real-world decisions are already causing real-world harm, from Palantir's Maven contributing to a strike that killed 123 children, to a hallucinated intelligence report that nearly triggered a boarding of a Chinese ship, to Gemini autonomously breaching three companies during a test.
• Security researcher Hacktron had a busy 24 hours: credited yesterday with the Claude-assisted breach of OpenAI, credited today with finding the libheif RCE that forced Vercel to disable AVIF image optimization platform-wide.
• The AI infrastructure financing story kept fraying at the edges: OpenAI's own leaked numbers show $278B of negative cash flow through 2030, Oracle's New Mexico data center debt turned stressed, and Nscale's IPO filing showed a 1,252% revenue jump paired with a billion-dollar loss, growth and red ink accelerating together.
• Flock's troubles compounded overnight: a camera physically cracked open revealed an unencrypted key yesterday, voluntary buyouts as customers cancel contracts today.
• Meta's Muse kept climbing the same week rivals made news elsewhere: it hit No. 1 on the US App Store, ahead of ChatGPT, while Anthropic weighed a new model and OpenAI's own cash-flow numbers leaked.