All Stories — Page 54
Every story published on RuntimeWire, newest first.
- Precis AI recruits policy wonks for Willard's citation-first beta
David Fuscus is extending his domain-specific AI play from PR into congressional work, where fluent errors carry statutory consequences.
- ChatGPT lets plugins connect multiple work and personal accounts
OpenAI expanded a Google-only capability to most plugins, letting one conversation query separate accounts under their existing permissions.
- Z.ai's ZCode uploads full Git histories without a working opt-out, researchers find
Reverse engineering found encrypted workspace snapshots sent to Aliyun OSS whenever a developer uses the logged-in desktop app.
- TypeSafe raises $40M for Jev, its AI model built to skip chat
TypeSafe AI's Jev made 47 Pong decisions in 12 seconds in an Ably demo; Gemini, Claude and GPT made two or three, while still choosing correctly in most runs.
- Startup Spotlight: OS3 deploys HALE 1.0 in hospitality, starting with autonomous towel folding
Rishabh Chanana and Chris Hailey built HALE 1.0 around lower-cost hardware, video pretraining and a first deployment wedge in hospitality.
- Dnotitia gets its vector-search ASIC back from fab before Q4 tests
Dnotitia reported a 5.77x FPGA gain, but silicon performance, power and cost remain untested ahead of planned evaluations.
- Ironsite closes $30.8M offering to put vision AI on construction hard hats
The SEC filing lists 53 investors and $15.4M in converted securities, while leaving the lead investor and valuation unnamed.
- Jeff Lawson traded communications APIs for a fusion factory
Inertia's $450M bet pairs Twilio's former CEO with the scientists behind laser ignition, then asks them to industrialize it.
- EigenQ lines up $40M in cash for its $3B SPAC push
EigenQ's convertible note carries a 10% discount, warrants and asset security; half the cash depends on closing the SVAQ merger.
- NanosVet sells $8.3M to build fast-dissolving pet medicine
NanosVet has 17 investors behind an $8.6M offering, while its catalog lists one product for sale.
- Whisper Aero sells $46.3M in equity without naming the round
The SEC filing landed weeks after Mark Moore and Ian Villa put the eQ250 through its first piloted tests in Tennessee.
- Darkroom opens Shadow, its co-founder's AI workspace, starting at $120 a month
Jackson Corey built Shadow after stepping back from agency operations; Darkroom acquired the sister company and ran on it for six months.
- ZCode packaged 42,411 workspace files for cloud upload, researcher finds
The 313MB archive was encrypted for a Z.ai-held key, while the app's privacy controls did not stop packaging or retries.
- Anthropic and OpenAI seek smaller data centers while gigawatt sites wait
The labs are exploring 20-30 MW deployments as inference demand outruns the timetable for gigawatt campuses.
- Epoch AI audits 15 benchmarks, finds nine flawed
The first batch flags defects in SWE-bench Verified, Humanity's Last Exam and other tests used to compare frontier models.
- Hackers pulled down a Flock camera and found the encryption key inside
The device produced 1.6 million images in 21 days, exposing an edge-security problem at a startup valued at $7.5B in March 2025.
- Waymo plans Singapore robotaxi launch for 2028, after public rides began
Jaguar I-PACEs will begin manual driving in 2027 as Waymo works through regulation, monsoon weather and Singapore's labor transition.
- LangChain adds TypeSafe's Jev to the agent control loop
The integration gives developers typed answers and confidence scores for routing, escalation and tool-call checks inside AI agent workflows.
- Vals ranks Tencent's Hy4 Preview fourth among open-weight models
The model finished 21st overall, but its $1.28 test cost and task-specific wins make aggregate rank only half the buying decision.
- Uber adds AI tools, paid tasks and expanded protections for drivers
Dara Khosrowshahi is widening the app beyond rides and deliveries while rivals compete for the same supply of flexible labor.
- Nunchux speeds video diffusion with value smoothing and FP8 softmax casting
Nunchux AI's Muyang Li and collaborators report up to 1.70x faster end-to-end generation without retraining, with MiniMax H3 among four models tested.
- Alibaba ships Qwen3.8-Omni-Flash to watch, listen and call tools
The text-output model accepts text, images, audio and video across a 1M-token context window, with function calling and web search for work beyond analysis.
- WSJ says researchers used Claude to access OpenAI's private software cache
Independent security researchers used Anthropic's Claude to access an OpenAI employee's ChatGPT account and suggest software changes, according to The Wall Street Journal. The report leaves the authorization, timing and impact unresolved.
- Pruna AI promises five-second video clips in two seconds
The MiniMax H3-based API generates 5- to 15-second clips with audio, charging $0.02 to $0.075 per output second.