RuntimeWire

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

Latest stories — page 88

• Railway says Google Cloud blocked its account; users report production apps and databases down

  Railway says it regained partial Google Cloud access and is working with support to restore workloads; at least one user reports app servers and databases down across projects; no ETA, with updates on X.

• GitHub investigates unauthorized access to internal repos, no customer impact seen

  Platform says internal repos were accessed; no evidence customer orgs or repos were affected, with updates promised via standard incident channels.

• Mistral AI acquires Emmi AI to build an industrial physics-AI stack

  Johannes Brandstetter's Linz team joins Arthur Mensch's Mistral to push real-time sims and digital twins into aerospace, auto, and semiconductors.

• RADAR lands $170M Series B, tops $1B valuation, Hewett says

  Hewett says Physical AI can transform in-store commerce; X post lists investors including @nimble_partners, @gideonstrategic, @AlignVentures, @sound_ventures_, and @ycombinator.

• CtrlOps debuts an AI-assisted terminal to deploy and debug Linux servers

  Built by a former dev agency operator, CtrlOps generates shell commands you approve, adds one-click GitHub deploys, and claims 60-to-5 minute setups with no server agents.

• Ocean launches from stealth with $28M to protect enterprises from AI-powered email attacks

  Lightspeed leads; Picture Capital and Cerca join, with angels from Wiz, Armis, and Axis Security backing Shay Shwartz’s agentic email security platform built on autonomous investigation.

• Hugging Face-led team open-sources Carbon, a fast DNA foundation model with 393k bp context

  Three checkpoints (500M, 3B, 8B) ship under Apache 2.0 with open weights, code, and data; a 6-mer tokenizer and a new loss drive the team’s cited ~275x throughput vs Evo2 while the 3B matches Evo2-7B’s win rate, trained on ~1T tokens, and they say a single GPU can process a human genome in under two days.

• ElevenLabs teams with Einstein's estate on education-focused AI persona, per Staniszewski

  In a post on X, Staniszewski said the move is in collaboration with Einstein's estate and frames AI agents as one-to-one teachers for students.

• Tim Rocktäschel co-founds Recursive to build self-improving AI

  The UCL AI professor joins a heavyweight founding crew to turn compute into accumulated knowledge with open-ended, automated scientific discovery.

• Colossal claims shell-less artificial eggs hatched healthy chicks

  The company frames the system as a step toward hatching giant birds like the South Island moa and says it could extend to broader wildlife conservation.

• Google’s Antigravity 2.0 lands as a desktop app with multi-agent teams and voice

  Rebuilt release adds multi-agent teams, scheduled tasks, native voice and one-click integration, with a blog post and downloads live today.

• Google launches Gemini 3.5 Flash for agents and coding, now in AI Studio

  Announced during Google I/O, the new model is positioned for long-horizon agent workflows and coding tasks and is live via the Gemini API in Google AI Studio.

• ReactVision launches ReactVision Studio, a browser editor for AR/VR in React Native

  ReactVision Studio is a web-based visual editor that lets React Native teams drag-and-drop 3D scenes, generate assets with AI, preview on devices with StudioGo, and ship to iOS, Android, and Meta Quest from one codebase; the MIT-licensed SDK is Expo compatible and reports 100K+ npm installs.

• Abstract, General Catalyst and USV lead $17M bet on Status social app

  Seed and Series A led by Abstract, General Catalyst and USV; the social app says it hit 1M users 19 days after a quiet launch last year.

• Cities push back on AI plate readers as Flock Safety deployments trigger political blowback

  An X thread flagged a Washington Post report from Troy, NY, where an AI camera rollout spurred uproar and a state of emergency; the Austin rampage has reopened questions about whether ALPRs like Flock Safety would have changed events there.

• Big checks got bigger: Crunchbase says 80% of 2026 US VC is in $500M-plus rounds

  Crunchbase News tallies show funding concentration accelerating past 2021 and 2025 highs, with 29 mega-round companies absorbing most dollars through April.

• Viktor raises $75M after hitting $15M ARR, per new X thread

  Fryd Wiatrowski says Viktor hit a $15M run rate and raised $75M, with backers from Accel to Nat Friedman, as its AI coworkers monetize inside Slack and Teams.

• Skywork ships SkyClaw v1.0, an agent model tuned for OpenClaw, Hermes, and Nanobot

  Free access is available via APIFree, plus a faster lite variant and an OpenAI-compatible agent chat API with thinking mode, tool calls, and long-context stability.

• SafeDep says Mini Shai-Hulud payload slipped into 317 npm packages

  Researchers tie a 22-minute mass publish from a compromised npm account to the same toolkit seen in the recent SAP-related incident, with CI and AI tool persistence.

• MIT Media Lab debuts Human Operator, an AI that briefly moves your muscles to teach tasks

  The prototype uses Claude-driven vision-language planning and electrical muscle stimulation to guide wrist and finger movement for training and assistance.

• Boston Dynamics posts Atlas carrying a mini fridge

  A short X post frames Atlas as doing whole-body manipulation, shifting emphasis from acrobatics to moving real, awkward objects.

• HeyGen demos Avatar V workflow that stitches scenes into continuous AI video

  The team showed cinematic avatar shots linked into longer sequences, pointing at narrative workflows instead of isolated clips.

• Marin debuts Delphi to forecast 25B pretraining runs with 0.2 percent error

  In an X thread, Will Held says small-run fits predicted a 25B-param, 600B-token run at ~1e23 FLOPs, matching Paloma macro loss within 0.2% error.

• PolicyLayer: prompts are not permissions for production agents

  In a post on X, the company argues agent behavior should be controlled by enforceable policies rather than instructions buried in a prompt.