AI Briefing: May 29 – 30, 2026

Generated: 2026-05-30 08:06 UTC | Coverage: May 28 – May 30, 2026 (48 hours)

🚨 Breaking (last 24h)

Hermes Agent v0.15.1 & v0.15.2 — Same-Day Patches for "The Velocity Release"

Hermes Agent shipped two rapid follow-ups to its massive v0.15.0 release (May 28).

v0.15.1 (GitHub): A same-day hotfix addressing the dashboard infinite-reload loop affecting Docker, hosted Hermes, and fresh installs in loopback mode. Also fixes Kanban worker SIGTERM handling, unifies the /model picker, adds /yolo session bypass, and resolves emoji rendering in the 19,932-entry skill registry. 28 commits, 21 merged PRs, 9 contributors.
v0.15.2 (GitHub): Packaging fix ensuring bundled plugin.yaml manifests ship correctly in wheel and sdist distributions.

Source: Hermes Agent v0.15.1 Release Notes | v0.15.2 Release Notes

OpenClaw v2026.5.28-beta Series — Three Betas in One Day

OpenClaw released three beta builds on May 29 (beta.2, beta.3, beta.4), iterating on agent runtime recovery and channel delivery safety:

Agent/Codex runtime recovery: Subagents now keep cwd/workspace separation, hook context stays prompt-local, session locks release on timeout abort, stale restart continuations are avoided, and Codex app-server/helper failures no longer tear down shared runtime state.
Channel delivery fixes: Safer outbound plugin hooks, Matrix room IDs, iMessage reactions/approvals, Slack final replies, Discord reconnections.
Provider coverage: OpenRouter o4-mini routing, xAI grok-3-beta image support, Stability AI stable-image-core.

Source: OpenClaw v2026.5.28-beta.4 Release Notes

OpenAI Rosalind Biodefense Product (May 29)

OpenAI launched Rosalind, a biodefense product aimed at strengthening societal resilience against biological threats. Details emerged via the OpenAI newsroom on May 29.

Source: OpenAI: Strengthening Societal Resilience with Rosalind Biodefense

📊 Market Moves (last 48h)

Anthropic Series H — $65B at $965B Valuation

Anthropic raised $65 billion in Series H funding, led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital, with participation from Capital Group, Coatue, D1, GIC, ICONIQ, XN, and others. The company reports $47 billion run-rate revenue and notes the round includes $15B of previously committed hyperscaler investments (including $5B from Amazon). Strategic infrastructure partners include Micron and Samsung.

Source: Anthropic Newsroom | Twitter commentary

Anthropic Opens Milan Office (May 27)

Anthropic announced its sixth European office in Milan, following London, Dublin, Paris, Zurich, and Munich. The office opening closely follows Pope Leo XIV's AI encyclical Magnifica Humanitas, which Anthropic co-founder Chris Olah addressed at the Vatican. Italian enterprise customers include Generali Group, Unipol Group, Angelini Pharma, Bracco Group, Enel Group, and Pirelli.

Source: Anthropic Milan Office

🔬 Research (last 48h)

arXiv Status

The arXiv API returned rate-limit errors during this briefing window (HTTP 429). Based on the most recent successful fetch (May 29), the latest batch was dated May 27, 2026. The 1–2 day indexing lag is normal. No new papers were captured in the 48-hour window due to API rate limiting.

Source: arXiv API

OpenAI Trustworthy Third-Party Evaluations Playbook (May 29)

OpenAI published "A shared playbook for trustworthy third party evaluations", sharing lessons on designing evaluations for frontier models. The post emphasizes that modern frontier models act as agents (using tools, tracking state across steps), which means evaluations must account for the "harness" — the environment and setup facilitating actions — not just the model output.

Source: OpenAI Blog

🛠️ Tools (last 48h)

Claude Opus 4.8 — Dynamic Workflows & Fast Mode

Anthropic shipped Claude Opus 4.8 on May 28, building on 4.7 with:

Dynamic Workflows: Claude Code can now tackle very large-scale problems by spawning parallel subagents.
Fast Mode: 2.5× speed, now 3× cheaper than previous models' fast mode.
Benchmarks: Coding and agentic task improvements over 4.7; the only model to complete every Super-Agent case.
Early tester feedback: Notably better judgment, asks right questions, catches its own mistakes, pushes back on unsound plans.

Source: Anthropic: Introducing Claude Opus 4.8

Anthropic SDK Python v0.105.2 (May 29)

Patch release with full changelog available on GitHub.

Source: GitHub Releases

💭 Industry Pulse (last 48h)

Twitter Highlights

@gdb (Greg Brockman, May 29): "codex for managing the codex UI" and "table of contents for your long-running chatgpt conversations" — signaling continued Codex productization beyond raw model access. (Twitter)
@JeffDean (May 29): Enjoyed a chat with Gemini leads @OriolVinyalsML and @NandoDFreitas, discussing continuous learning systems. (Twitter)
@goodfellow_ian (May 29): Promoted TechAide fundraiser organized by Hugo Larochelle. (Twitter)
@JulianGoldieSEO (May 29): Multiple tweets on Hermes v0.15: "Hermes v0.15 might be the biggest free AI agent update nobody is paying attention to. Faster startup. Agent swarms." (Twitter)
@GSuiee (May 29): Summarized Claude Opus 4.8 improvements: agent coding 64.3→69.2%, knowledge work 1753→1890, Dynamic Workflows enabling hundreds of parallel subagents. (Twitter)

Hacker News Front Page (May 29)

📡 Sources & Data Provenance

Source	Status	URL
Twitter/X API	✅ Working	https://twitterapi.io
GitHub Releases	✅ Working	https://github.com
arXiv API	⚠️ Degraded (HTTP 429)	https://arxiv.org
Wiki Raw Archive	✅ Available	~/wiki/raw/
Web Extract (jina.ai)	✅ Working	https://r.jina.ai
Hacker News	✅ Working	https://news.ycombinator.com

Each story above links directly to its primary source. Unlinked claims were cross-referenced from multiple sources.