Anthropic just shattered expectations, hitting a $1 trillion implied valuation on secondary markets — tripling from its $380B funding round in January — fueled by $30B annualized revenue and a 233% quarterly surge driven by Claude Code adoption, overtaking OpenAI on the secondary market for the first time. OpenAI fired back with GPT-5.5 now live in the API, its most capable agentic model yet, claiming top marks for multi-step coding and autonomous reasoning. The agent autonomy race became vivid through @steipete's Codex dispatches: rate-limited on GitHub, Codex opened Cloudflare, generated a new API key, navigated the browser UI, and closed the issue — no human in the loop. OpenClaw is at 347,000 GitHub stars but facing a growing security story: 13 CVEs in April, 135,000 exposed instances flagged by SecurityScorecard, and a consent-bypass exploit class (CVE-2026-41349) — essential reading if you're self-hosting.
Secondary market buyers are racing for a dwindling supply of Anthropic shares, pushing its implied valuation to $1T — nearly tripling from its $380B round just three months ago. Annualized revenue jumped from $9B at end-2025 to $30B by March 2026, a 233% surge driven by Claude Code adoption.
GPT-5.5 and GPT-5.5 Pro are live in the OpenAI API as of April 24. OpenAI bills it as its smartest agentic model — better at multi-step planning, computer use, and early-stage scientific research, with a 1M token context window and 82.7% on Terminal-Bench 2.0.
DeepSeek unveiled its next-generation DeepSeek-V4-Pro with a 75% developer discount through May 5. Market reaction has been more subdued than last year's shock debut, but the pricing pressure on western AI labs continues to intensify.
SecurityScorecard found ~135,000 exposed OpenClaw instances, ~50,000 directly exploitable. ClawHavoc — 1,184 malicious skills on ClawHub — is actively targeting users via credential stealers and clipboard hijackers. CVE-2026-41349 allows LLM agents to silently bypass execution consent via config patch. Update immediately if self-hosting.
The State Department has launched a global diplomatic push flagging systematic efforts by Chinese AI firms — including DeepSeek — to steal intellectual property from US AI labs. Allies are being asked to tighten access controls and share intelligence on exfiltration attempts.
Selected posts from the edition, with links to the originals.
"I'm again rate limited on GitHub, but codex just opened the browser and clicks around GitHub as workaround."
View post on X ↗"LATEST: ANTHROPIC HITS $1 TRILLION IMPLIED VALUATION"
View post on X ↗"Codex complained that one API key didn't had the permissions it needed, so told it to get a new one, it opened @Cloudflare and created a new API key."
View post on X ↗"Yes it did type into the comment box and closed the issue."
View post on X ↗retweeting @openclaw: "OpenClaw 2026.4.26 🦞 — Google Live Talk, Better Ollama/local models, Bring over Claude + Hermes setups, One-command Matrix..."
View post on X ↗"GStack ships with Chromium that has Claude Code as the CLI and you can type /pair-agent from your real CC or the sidebar one — it'll give you a string to paste into your OpenClaw/Hermes so it can control your browser remotely, and you can log it into stuff and have it scrape"
View post on X ↗"@sama @filicroval For anyone who hit Claude Code's session limits and didn't want to pay $200, this is exactly the alternative they needed at exactly the right moment."
View post on X ↗"Google Nanobanana and Claude design users."
View post on X ↗"Remember Code Red? all that has happened is that OpenAI has been able to raise more and more money.. 'Exclusive!'"
View post on X ↗"A 23-year-old researcher and an undergraduate collaborator used the $200-per-month professional tier of ChatGPT to address a 60-year-old number theory challenge known as the Erdős conjecture on primitive sets. The AI provided a raw, unconventional logical framework that bypassed [conventional approaches]."
View post on X ↗