RuntimeWire

Real-time startup intelligence for the AI economy. Funding, model launches, infra shifts, and founder moves — the signals that matter.

Latest stories — page 71

• Jack Hidary's SandboxAQ wins $500 million CHIPS award for AI chipmaking materials

  The Commerce Department deal gives the U.S. a minority stake and pushes SandboxAQ from model vendor into industrial policy contractor.

• Anthropic's Mythos fight turns on the hacker who first told it to slow down

  Nicholas Carlini warned Anthropic not to release Mythos in March. He is now central to its case that guarded access is safer than a ban.

• AGI House turns its builder calendar toward agents and automation

  Rocky Yu's Hillsborough AI community is using build days to pull founders from model demos into live-web agents, identity, and enterprise workflows.

• Cursor Says Its Next Coding Model Is 1.5T Parameters and Training on 100,000 GPUs

  A Compile disclosure shows Cursor pushing beyond tool design into frontier-scale model ownership.

• Qualcomm's reported Tenstorrent talks value Jim Keller's AI chip team at $8B to $10B

  The reported approach would test whether Qualcomm wants another NUVIA-style team acquisition or a broader RISC-V AI architecture bet.

• Unicorn Engine's decade-long CPU emulation bet still has teeth

  Nguyen Anh Quynh and Dang Hoang Vu built Unicorn as a focused alternative to QEMU, and security tooling still leans on that choice.

• GrapheneOS ports to Android 17 on release day

  The privacy-focused Android fork says an initial Android 17 build is planned for June 17th across supported devices.

• Leaked OpenAI financials put a hard number on Altman's compute burn

  Audited documents viewed by Ed Zitron and verified by the FT show $13.07B in 2025 revenue and a $38.5B net loss attributable to OpenAI.

• Coinbase rolls out AI investment advisor to Coinbase One members

  The SEC-registered advisory product gives automated portfolio guidance while keeping trade approval and execution in the user's hands.

• Anthropic fight with Trump turns into a cyber-defense fight

  Security leaders say the Fable 5 and Mythos 5 restrictions risk punishing the same bug-finding work defenders need.

• SpaceX agrees to buy Cursor maker Anysphere for $60 billion

  The stock deal gives Elon Musk's public-market SpaceX a fast-growing AI coding business as xAI chases Anthropic, OpenAI and Microsoft.

• YC Names Christopher Golda and Grey Baker General Partners

  The appointments add two YC-linked operators to Garry Tan's senior bench as YC runs more batches and competes earlier for technical founders.

• SubQ Releases Its 1.1 Small Model Card as Dangel and Whedon Try to Prove Long Context Can Beat RAG

  The Subquadratic team says its sparse-attention model hits 98% needle retrieval at 12M tokens, but access remains limited to design partners.

• Genesis AI's Eno robot rejects the humanoid default

  Xian Zhou and Theophile Gervet are taking a wheeled, foldable path into physical AI after raising a $105 million seed round.

• Analysis: Why Salesforce is buying Fin for about $3.6B

  Salesforce + Fin: packaging customer-service agents for Agentforce

• Moonshot AI's Yang Zhilin Pushes Kimi Deeper Into Coding Agents

  Kimi K2.7-Code is a 1T-parameter MoE model with 32B active parameters, a 256K context window and open weights on Hugging Face.

• Anthropic faces class-action claim over Claude Max 20x limits

  Karl Kahn says Claude Max 20x delivered six to eight times Pro usage, not the 20x Anthropic advertised.

• Boz tells Meta employees its AI reorg rollout was 'atrocious'

  Meta is trying to repair morale inside a 6,500-person AI unit formed in March after employees pushed back on forced transfers and lower-status work.

• Microsoft's GitHub capacity crunch sends it to AWS

  AI coding agents have turned GitHub reliability into an infrastructure problem Azure cannot absorb alone on Microsoft's timetable.

• Hermes Agent's new async subagents take aim at the blocking-agent problem

  Teknium says the delegate tool can now fan out work without freezing the chat, a practical change for long-running agent workflows.

• Cartesia packages Sonic-3.5 and Ink-2 into a full voice-agent stack

  Karan Goel is using benchmark wins to pitch Cartesia as both the speaking and listening layer for real-time AI agents.

• CrankGPT is a hand-cranked AI project with real edge-computing numbers

  CrankGPT runs speech recognition, a small language model, and text-to-speech locally on a Raspberry Pi 5 with no battery or cloud.

• Claude Code user says the coding assistant saved his life by pushing him to the ER for AFib

  A 73-year-old developer said he mentioned feeling unwell during a coding task, and Claude Code kept urging immediate care before doctors treated a sudden AFib episode.

• SGLang adds DFlash to push Qwen 3.5 397B-A17B inference up to 4.3x faster

  Z Lab, Modal and LMSYS released a DFlash drafter for Qwen's 397B model and benchmarked it above native MTP on 8x B200 GPUs.