OpenAI and Anthropic public-post watch

Checked 19 August 2026 at approximately 3:10 am AEST (5:10 pm UTC on 18 August). Priority window: the preceding 24 hours. The most recent recorded AI bulletin was 17 August, so one important later post outside the strict window is included and labelled.

Headlines

• OpenAI and NVIDIA are planning a 4.25-gigawatt AI campus in Ohio — Sam Altman said he was excited to work with NVIDIA after Jensen Huang published the terms of the PORTS-Pike project. NVIDIA says it will work with SB Energy to secure the land, power, and data-centre shell, with OpenAI as tenant and operator of an AI factory built on NVIDIA’s DSX platform. The initial deployment is expected to provide 4.25 GW, come online in phases from 2028 to 2030, and could accommodate roughly 1.5 million GPUs per hardware generation. NVIDIA says OpenAI’s existing and planned commitments amount to about 12 GW of NVIDIA compute through 2030, potentially rising to 16 GW if another 3.75 GW at the site is secured; it values that broader opportunity at roughly $600 billion. These are NVIDIA’s long-range estimates and contractual description, not delivered capacity or recognized revenue. NVIDIA also says its support covers defined lease, power, and residual-value obligations rather than the project’s entire cost. Altman’s post · Huang’s original article

• Claude Code added an editable UI-design workflow in research preview — Anthropic’s Boris Cherny highlighted the new /design skill and asked users for feedback. The preview brings Claude Design’s artifact-based artboard workflow into Claude Code CLI and Desktop: users can generate multiple editable UI artboards, choose and refine one, then ask Claude to implement it. The announcement calls this a research preview, so it should not be read as a promise of production stability or universal availability. Cherny’s post · Claude Developers announcement

• Claude Code cut high-percentile CPU use by half — Cherny also amplified an implementation change that Anthropic says makes the CLI use two times less CPU at p99. The team traced the spikes to Bun’s garbage collector running on a fixed timer and changed it to wait until the process is idle instead of competing with an active turn. This is a p99 CPU measurement, not a claim that every Claude Code task now completes twice as fast. Cherny’s post · Claude Developers explanation

Important update since the last recorded bulletin

• Codex’s optional one-million-token Sol context now works with ChatGPT-account usage — Tibo Sottiaux said the previously documented configuration, which had only worked with API keys, was enabled for Codex sessions billed through ChatGPT accounts. His example selects GPT-5.6 Sol, sets a 1,000,000-token context budget, and starts automatic compaction around 900,000 tokens; he notes that Sol’s documented window is 1,050,000 tokens. Sottiaux and Noam Brown both cautioned that OpenAI tuned the smaller default and its automatic compaction carefully for performance and cost, so this is an opt-in escape hatch rather than a new recommended default. The account-support switch was posted on 17 August, after the previous bulletin but outside today’s strict 24-hour window. Sottiaux’s account-support update · Configuration post · Brown’s context

Other activity

• OpenAI is openly soliciting “obvious” missing product ideas — Within the priority window, Sottiaux asked users what reachable improvement OpenAI should already have made across Codex, the API, or its models. This is a product-feedback request, not a roadmap commitment or feature announcement. Original post

• Boaz Barak cautioned against treating AI monitoring as a complete safety solution — Responding to concerns about monitoring failures, Barak compared monitors to seatbelts: valuable safeguards that do not remove the need to operate safely. The brief post is a safety principle from an OpenAI researcher, not a newly announced company policy. It appeared after the prior bulletin but outside today’s priority window. Original post

Steve Yegge watch

• No fresh visible main-feed post in the priority window — Yegge’s newest visible main-feed items were posted on 17 August: a companion comic titled “Fable Likes an Idea” and a watermarking joke. Neither conveyed a material product, research, policy, or company update, so they are not promoted to the headlines. Fable post · Watermarking joke

Coverage note

Live public timelines were checked across notable OpenAI leadership, product, developer, research, and safety staff including Sam Altman, Greg Brockman, Romain Huet, Tibo Sottiaux, Adam Fry, Eric Provencher, Noam Brown, and Boaz Barak. Anthropic checks included Dario Amodei, Boris Cherny, Alex Albert, Jack Clark, and Mike Krieger, with the fresh Claude Code announcements cross-checked against the public Claude Developers posts. Steve Yegge’s profile was reviewed separately. Previously covered token-price comparisons, Dario Amodei’s regulation and biology thread, ChatGPT’s 14 August feature drop, cross-model delegation, Ultrafast mode, content marking, cybersecurity access, usage resets, and auto mode were not repeated.