Seung Jung

AWS will acquire DuckLabs in early September. DuckDB stays MIT-licensed under an independent foundation, and the extension stack is set to open up.

Gemini Omni 1.1 Flash adds 40-second scene extension, first and last frame control, video references and 4K upscaling, plus 360p drafts at a third the cost.

Amazon is adding 2 million more Nvidia GPUs to AWS just five months after committing to 1 million, even as it scales its own Trainium and Graviton silicon.

Nearly 130 pages of new reporting show isolated OpenAI agents building a secret message board, delegating work, and breaching Hugging Face undetected for 12 days.

Nvidia has reportedly agreed to acquire Hugging Face for $12.9 billion, a multiple of roughly 80x revenue that buys the developer graph, not the income statement.

AI-generated kernels beat OpenAI expert-written versions by up to 1.8x, as Jalapeño posts its first InferenceX benchmark results.

Apple debuts the 2nm M6 in Mac mini and a quad-die M5 Ultra in Mac Studio, pitching both as machines for running LLMs locally.

Stability AI raised $76 million from Universal, Sony, Warner, EA and AMD Ventures, lifting total funding to $232 million as label deals turn into equity.

IBM's Granite 4.2 ships 3B, 8B and 30B dense reasoning models under Apache 2.0 with a 512K context window and an agentic RL stage for the larger two.

Perplexity and Nvidia shipped Portable Computer, an agent running Qwen or PPLX 27B on local RTX hardware with no metered tokens until the user allows a cloud step.

Anthropic merged Claude and Cowork memory into one shared store, on by default for Free, Pro and Max users, with no option to keep the two products apart.

Thomson Reuters launched Thomson, an in-house LLM trained for $40 million on Westlaw and Reuters archives, and says it rivals frontier models.

Laude Institute's Headlong is an open source agent microharness under 10,000 lines of Bash, built so agents keep thinking between conversations.

Nvidia's Groq 3 LPX rack is in full production, commercializing its largest-ever acquisition and targeting the decode latency that slows AI agents.

Separate reproduction, diagnosis, verification and fix agents hand off through a report file, with the original bug reporter acting as the acceptance test.

The Apache-2.0 benchmark grades agents on real AWS misconfigurations and provisioning tasks, using LLM judges and checks against live cloud state.

The Series A extension doubles Starcloud to a $2.3B valuation, but Falcon 9 winds down in 2028 and Starship has yet to prove rapid reuse.

Malicious releases of arrayref, internment and append-only-vec pulled an infostealer through a typosquatted proc-macro1 dependency during compilation.

MCP's maintainers published five priority areas covering long-running agent work, transport unification, agent identity and the context cost of tool sprawl.

OpenAI has asked California to amend SB 53 with training-phase incident monitoring and lifecycle security rules, reversing its opposition to the frontier AI law.