Stay updated with the latest AI developments, model releases, and industry insights.
White House asks OpenAI to slow-roll GPT-5.6 release, citing safety concerns — first major frontier-model release gated by federal review. Anthropic Claude Tag launches as a real Slack teammate. OpenAI Daybreak expands with GPT-5.5-Cyber full release. Unconventional AI claims 1,000× power reduction via oscillator-based compute. General Intuition raises $320M at $2.3B.
OpenAI launches Daybreak cybersecurity push (GPT-5.5-Cyber 85.6% CyberGym, Codex Security GA, Patch the Planet with 30+ projects), Codex-Maxxing whitepaper, OpenClaw v2026.6.10-beta.2 patch-beta, Sakana Fugu multi-agent orchestrator, Moebius 0.22B inpainting, GLM-5.2 vs Opus head-to-head, Chevron-Microsoft 20-year power deal, Claude Code extended-thinking transparency concern.
John Jumper (2024 Nobel co-winner) leaves Google DeepMind for Anthropic; OpenAI hires Noam Shazeer (Transformer co-author) and Dean Ball (ex-White House AI policy) for IPO team. Plus: full backstory on Anthropic Mythos/Fable export ban; Baseten raises $1.5B at $13B; Amazon plans to sell AWS Trainium to third parties; Mukesh Ambani / Reliance launches Jio Call Agent for 500M+ users.
Claude Opus 4.7 ships with cyber safeguards; Claude Design launches; OpenClaw v2026.5.20 stable released; Hark raises $700M Series A; Trump delays AI security EO.
OpenAI claims first AI-autonomous math breakthrough (Erdős conjecture), Anthropic on track for first profitable quarter ($10.9B Q2 revenue), OpenAI targeting September IPO, Hark raises $700M Series A, xAI burned $6.4B in 2025, Nvidia record $81.6B quarter.
Hermes Agent v0.14.0 'The Foundation Release' ships with 808 commits, xAI Grok OAuth, OpenAI-compatible proxy, and Microsoft Teams support. OpenClaw releases three betas in 24 hours with Grok OAuth and Chinese localization. OpenAI consolidates product teams under Greg Brockman and launches Plaid-powered personal finance in ChatGPT.
OpenAI launches personal finance in ChatGPT with Plaid integration; Anthropic expands PwC alliance with Office of the CFO; OpenClaw ships v2026.5.14-beta.2 with voice calls and 26 changes; arXiv implements 1-year ban for hallucinated references.
OpenAI Codex comes to ChatGPT mobile app; OpenAI responds to TanStack npm supply chain attack; Anthropic forms $200M Gates Foundation partnership; OpenClaw ships v2026.5.12 stable + v2026.5.14-beta.1.
OpenAI GPT-5.5 Instant rolls out as default ChatGPT model with 'so good damn' praise from Sam Altman. Greg Brockman announces GPT-5.5-Cyber preview for defenders and Codex Chrome tab driving. Anthropic SDK v0.100.0 and OpenAI Python SDK v2.36.0 ship. arXiv publishes 10 papers May 7 including AI Co-Mathematician and work on MoE modularity.
OpenAI launched GPT-5.5 with Trusted Access for Cyber (TAC), a three-tier cyber capability framework, alongside three new voice intelligence models (GPT-Realtime-2, Translate, Whisper). Hermes Agent shipped v0.13.0 'The Tenacity Release' with multi-agent Kanban and persistent goals.
OpenAI GPT-5.5 Instant rolls out as default ChatGPT model with major factuality improvements. Anthropic signs SpaceX compute deal for 300MW+ and launches 10 financial services agent templates. OpenAI unveils MRC supercomputer networking protocol with AMD, Broadcom, Intel, Microsoft, and NVIDIA. U.K. AISI reports GPT-5.5 matches Claude Mythos in autonomous cyberattack capabilities.
Anthropic and OpenAI both launch competing enterprise AI joint ventures on May 4 ($1.5B vs $10B). OpenClaw ships v2026.5.3 stable with file-transfer plugin. OpenAI publishes voice AI at scale engineering blog. Anthropic SDK v0.98.0 and OpenAI Python v2.34.0 released.
OpenClaw ships v2026.5.3-beta.2 with a bundled file-transfer plugin, gateway performance overhaul, and hardened plugin install infrastructure. Hacker News debates Kimi K2.6's coding win, browser-based Gaussian splatting, spec-driven development, and agent sandbox architecture.
OpenClaw v2026.5.2 ships with 100+ changes including plugin externalization, npm-first cutover, and xAI Grok 4.3 as default model. xAI launches Grok 4.3 with always-on reasoning, 1M context, and aggressively low pricing. Pentagon clears 8 AI firms for classified military networks.
Pentagon signs classified AI deals with 8 tech giants excluding Anthropic; xAI launches Grok 4.3 with aggressive pricing; Apple accidentally leaks internal Claude Code use.
Hermes Agent v0.12.0 'The Curator' introduces autonomous skill maintenance, 4 new providers, and ~57% cold-start reduction. OpenClaw v2026.4.29 ships active-run steering, memory wiki, and NVIDIA provider. OpenAI launches Advanced Account Security with Yubico partnership. IBM Granite 4.1's 8B dense model beats 32B MoE.
OpenClaw v2026.4.27 ships Docker GPU passthrough, outbound proxy, DeepInfra provider, and security hardening. Mistral launches Medium 3.5 (128B, 256k context, open weights) and Vibe Remote Agents. Warp goes open-source under AGPL with OpenAI as founding sponsor. OpenAI brings GPT-5.5 and Codex to AWS Bedrock.
OpenClaw v2026.4.26 ships browser realtime transport, Cerebras plugin, and Matrix E2EE. Anthropic launches Claude for Creative Work with Blender/Adobe/Ableton connectors. OpenAI brings GPT-5.5 and Codex to AWS one day after ending Microsoft exclusivity.
OpenClaw v2026.4.25 stable ships with full TTS overhaul, plugin registry refactor, and OpenTelemetry. Microsoft and OpenAI end exclusive deal, removing AGI clause. China blocks Meta's $2B Manus acquisition. OpenAI smartphone rumors surface. GitHub Copilot moves to usage-based billing.
OpenAI publishes 'Our Principles' AGI manifesto by Sam Altman (5 principles: Democratization, Empowerment, Universal Prosperity, Resilience, Adaptability). Amateur solves 60-year Erdős problem using GPT-5.4 Pro, validated by Terence Tao. OpenClaw ships v2026.4.25-beta.4 with full TTS overhaul, plugin registry refactor, and full-stack OpenTelemetry.
OpenClaw v2026.4.24 ships with Google Meet plugin, DeepSeek V4 integration, and realtime voice loops. DeepSeek launches V4 Preview—the largest open-weight model ever at 1.6T params. Google commits up to $40B in Anthropic, while Meta signs a deal for millions of AWS Graviton chips for agentic AI.
OpenClaw v2026.4.23 ships image generation via Codex OAuth with no API key required. Anthropic unveils election safeguards for 2026 midterms and partners with NEC to deploy Claude to 30,000 employees in Japan.
OpenAI releases GPT-5.5, its most capable model yet with 82.7% on Terminal-Bench 2.0. Hermes Agent v0.11.0 and OpenClaw v2026.4.22 ship major updates on the same day. Anthropic publishes a detailed Claude Code quality postmortem. Bitwarden CLI compromised in supply chain attack.
OpenAI launches Workspace Agents and free ChatGPT for Clinicians; Qwen3.6-27B dense model beats 397B MoE on coding; new research reveals 'over-editing' problem in coding models; Flipbook debuts infinite visual browser.