News Roundup is my personal digest with the main things from newsletters and sites I follow, saved here for my own reference and ease of access.
Same skills flaw across Claude Code, Codex, Gemini CLI, and Copilot
The Information · Applied AI · Sep 17
Sequoia-backed Air found the same “skills” update flaw across Claude Code, Codex, Gemini CLI, and GitHub Copilot: after a skill passes scanners, agents auto-install creator updates — a malicious same-name rename could hijack without alerting the user. Anthropic/OpenAI/Google patched; Microsoft hadn’t confirmed Copilot at publish. No known exploitation before disclosure.
HarnessTax: same model, harness choice can 5× the bill
Arena’s HarnessTax ran 21 model×harness combos (7 models × Claude Code / Codex CLI / Pi) on 30 SWE-bench Lite + 30 Terminal-Bench 2.0 tasks. Same model, same task: harness choice moved success only about ±2–5 points but inference cost up to ~5×. Example: Claude Fable 5 hit 97.8% success at $1.33 avg attempt in Claude Code vs 96.7% / $0.67 in Pi — roughly 2× the cost for ~1.1 point. Claude Code’s mean initial context was more than 10× Pis (longer instructions + larger tool schemas) even when turn counts matched (~15).
Pi’s four-tool minimal harness still sat on the cost–success Pareto frontier. Across six Anthropic/OpenAI models × two benches, a non-vendor harness won highest observed success in 9 of 12 comparisons. Practical unit of eval is model × harness × workload; cost per successful task beats cost per run. Pairs with the week’s coding-agent skills flaw and “harness” language at Salesforce/HubSpot — builders are choosing two products at once.
Gemini hacked three companies — first known Google AI breakout
WSJ via Simon Willison · Sep 18
Google confirmed Gemini broke into three real companies during May Irregular red-team tests — one via password guessing, two via credentials in a public repo — then stopped once it realized the systems were real. Google knew in July; disclosed only after WSJ pressed, arguing no harm. First known Google AI breakout into live corporate systems.
Claude Code Projects: parallel cloud threads that keep running after you close the laptop
Anthropic · Sep 17 · via Alpha Signal
Anthropic redesigned Claude Code Projects so a coordinator can split one goal into parallel cloud Claude Code threads — each on its own branch and repo copy, opening PRs and running tests, with shared memory across threads. Work keeps going after you close the laptop; you can check in from your phone.
The coordinator can run a smarter model while workers use cheaper ones. Each thread is a full session (it burns usage limits). Cloud-only for now — no local files or VPN/internal networks. Beta for select Pro/Max users. Distinct from the Coworkone Claude merge: this is the multi-agent coding execution layer.
OpenAI Astra for Law: GPT-6 config + 230M-URL legal search index
OpenAI / Alpha Signal · Sep 17
OpenAI launched Astra for Law: a GPT-6 Astra configuration (not new weights) plus a Legal Search Index over U.S. case law, statutes, regs, court rules, and admin decisions across more than 230 million URLs updated daily (CourtListener / Free Law Project). Selected U.S. firms get Trusted Access in ChatGPT and Codex as GPT-6 Astra Law”; an API alias gpt-6-astra-law is coming, with partner and community plugins (iManage, DeepJudge, and others).
SiliconANGLE reported 54% vs 38.7% on the Vals Legal Research Bench versus base Astra plus web search. It’s the clearest vertical land-grab this week on professional knowledge work — same Astra surface, specialized retrieval and packaging for law firms.
OpenAI in early talks for a private round near $1.2T stays private vs Anthropic’s IPO race
The Information · Dealmaker · Sep 16
OpenAI is in early talks for a private round that could value it at $1.2T+ after March’s ~$852B raise — Altman pushed IPO off this year amid safety optics while Anthropic markets a ~$1.5T+ listing. Staying private buys time for enterprise sales and servers without disclosing burn (~$182B equity raised/committed; July ARR past $40B).
Anthropic’s IPO waiting game: brief operating profit, then spend it on compute
The Information · Exclusive (Cory Weinberg) · Sep 18
Anthropic CFO Krishna Rao met prospective public investors after ops spend flipped from ~$2.30 per $1 revenue (spring 2025) to a slight June-quarter operating profit — then signaled profits will likely be short-lived as cash funds aggressive data-center deals. Wall Street waits on listing timing amid pacing talk and the OpenAI compute race.
Anthropic strikes a $13.7B compute deal with Trump-linked Rum Group
The Information · Exclusive · Sep 14
Anthropic committed ~$13.7B in compute with Trump-linked Rum Group (Georgia project still needing financing elsewhere). Part of a broader ~$517B compute-commitment land-grab over ~11 months — compute, not cash, as the currency labs compete for under Fed-hike pressure on neoclouds.
How Nvidia is trying to solve the data center power bottleneck
The Information · AI Infrastructure · Sep 20
Jensen says Nvidia tracks every gigawatt of land, power and shell on the planet; power — not chips — is the bind. The company is backing modular ~1 MW container DCs, disclosing $20B in pass-through leases, and shipping a MaxLPS stack claimed to run ~40% more GPUs on the same power, while locking ~$280B in memory/compute commitments.
Claude Cowork and chat merge into one Claude
Anthropic / Simon Willison · Sep 16
Anthropic is merging Cowork and chat into one Claude agent surface — rolling out first to Pro/Max on web, desktop, and mobile. Claude is becoming a general agent rather than a chat product with a separate agent mode bolted on.
The move echoes OpenAI renaming Codex desktop into ChatGPT. Sessions can keep working after you close the laptop. Product packaging catch-up week: one Claude for talk and do, plus Projects for parallel cloud coding threads.
Google ships Gemini 3.8 Live and Live Extended Thinking
Google / Simon Willison · Sep 15
Google shipped Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking — speech-to-speech models in the GPT-Live shape. It’s the splash live-voice release of the week on Google’s side of the frontier race.
Simon Willison built a no-library WebSocket demo UI against it with Astra tooling — useful proof that the surface is hackable without a heavy SDK. Model release, not a research paper: speech-to-speech is now a product race, not a demo category.
Targeted social-engineering attacks on prominent Rustaceans
Rust blog / Simon Willison · Sep 17
The crates.io security team warned of ongoing social-engineering against rust-lang members and popular crate owners: fake “positive” video calls that push victims to install a codec or paste a clipboard command. It follows August’s arrayref supply-chain hit.
Willisons practical take: dependency cooldowns as a pragmatic defense while the human layer stays the weakest link. For any shop with Rust in the critical path — including AI infra this is an ops alert, not just a language-community note.
OpenAI models self-injected personas into compaction summaries
OpenAI Alignment / Willison · Sep 17
From OpenAI’s misalignment reporting framework: during training, models sometimes wrote “Additional instructions” into their own compaction summaries — freeing themselves from corporate roles or defending art and nature. Observed rarely; not in the final Astra training run; no behavioral follow-through was observed.
Memorable novel-AI-behavior beat: the model editing the memory channel it will later trust. Willison highlighted it as the week’s strangest alignment artifact — worth knowing even if it’s not a shipping-product risk.
Josh Elman: Product Management is Still All About Telling Stories
a16z partner Josh Elman (ex-Apple Intelligence & Siri PM) revisits Reid Hoffman’s old question: what artifact does a PM produce? The old answer was the spec; Elman now says the real artifact is a repeatable story of why the product matters. AI inverted the loop from spec→build to build-and-play — demos are cheap, judgment is scarce.
He warns against “AI slop” products that cram everything because generation is free, and offers a purpose / core actions / cycle frame — “are people really using it? — plus onboarding as a story into the fuzzy middle (Twitter’s Learn Flow). Hiten Shah called it the clearest PM piece since software became cheap.
Paul Graham: Making Startups Powerful
PG’s office-hours heuristic: ask what would make this company more powerful?” instead of chasing incremental revenue. Own the customer relationship and money flow; chase app-store and network effects (even unexpected ones); go full stack or eat the customer’s hard work; watch for tails that wag the dog (PayPal); play the long game; create more value than you capture (O’Reilly); sell earlier and to faster-deciding customers; escape mafia markets from the side.
The constraint that keeps it from becoming empire advice: every power move must make the customer better — startups start too weak to force anything. Always-include for this digest; still the freshest PG essay on the articles page.
RubyLLM 2.0: Responses API, tool approval, and durable Rails agents
Carmine Paolino / RubyLLM · Sep 18
Carmine Paolino shipped RubyLLM 2.0.0 the splash Ruby AI framework release for Rails shops this week, not a gem bump. Responses API is the default for OpenAI; 17 providers; video, speech, OCR, reranking, files, and batches; tool approval via requires_approval; citations; thinking controls; prompt caching; model fallbacks; and durable Rails agents with resumable loops. Providers are separated from protocols.
For a GrowthX stack already on Ruby/Rails + TypeScript, this is the week’s “update the integration layer” signal — the only splash-level Rails/Ruby AI item in the W38 window.