All Stories — Page 63
Every story published on RuntimeWire, newest first.
- Cognition brings Dioxus team into Devin's runtime work after $2B raise
Dioxus will stay open source under Cognition's support while Jonathan Kelley and his engineers work on Devin's virtual machines, computer use, and testing.
- Enigmata raises $6.5M to run AI without exposing plaintext
Scott Searle's Enigmata, founded in 2024, says Cipher keeps enterprise data encrypted during training, search and analytics; Blockchange Ventures led the seed.
- Cursor launches Projects, a persistent manager for fleets of coding agents
Michael Truell's February thesis becomes a beta that remembers months of work, watches Slack and PRs, and keeps coding after the laptop closes.
- Microsoft reportedly targets 38 GW of data-center capacity by 2032
The planning target would more than triple Microsoft's current capacity, with about one-third devoted to AI infrastructure as calendar 2026 capital expenditures are expected to reach $175 billion.
- Baseten announces Blaxel acquisition to put agent runtimes beside model inference
Baseten says the acquisition will add sandboxes, persistent storage and networking for long-running agents, with Blaxel remaining available during the integration.
- Fluidstack is in talks for a $5B Pentagon loan to build AI data centers
The proposed financing would give Gary Wu and Jamie Cox another source of capital as Fluidstack builds data centers for AI labs and secures power for further expansion.
- xAI is changing the economics of X Search. RuntimeWire was built for a narrower retrieval model
Starting September 21st, xAI will stop charging a flat fee for Grok’s X Search calls and begin billing for every post and user profile returned. For products that depend on large-scale X ingestion, the difference could be significant. RuntimeWire was architected around a different assumption: use human editorial judgment to narrow the field first, then let machines investigate deeply.
- Elly pairs recruiters with AI on every search, promising faster hiring at half the cost
CEO Kristen Habacht is turning lessons from Trello, Atlassian and Typeform into a recruiting service built around software and human judgment.
- Anthropic's Claude community starts global Fable 5.1 Build Days
Claude Community organizers will hold buildathons from Austin to Melbourne between September 11th and 25, inviting participants to bring problems, ideas or unfinished projects.
- Google opens Dreambeans' AI story feed to all US adults
The Labs app turns connected Gmail, Calendar, Photos, YouTube, Search and Gemini data into illustrated daily recommendations.
- OpenAI launches Agents API to run the Codex harness as a managed service
The public beta manages context and subagents while letting developers use OpenAI, partner, or self-hosted sandboxes.
- Summation launches a $60 AI analyst that promises to check its work
Ian Wong is pushing the $35M-backed enterprise platform into self-serve software with shared context, traceable answers and scheduled workflows.
- ChatGPT puts Dropbox, Box and SharePoint files directly inside Library
The web rollout lets paid users browse cloud files, attach them by @mention and work across folders in both Chat and Work.
- MiniMax joins Nebius builder program without disclosing a cloud deal
MiniMax says the program will give developers access to models, infrastructure, tooling, credits and office hours, but neither side disclosed a commercial contract or production deployment.
- OpenAI will pause $200 Pro subscriptions as Astra strains capacity
Existing subscribers keep access, while OpenAI leaves its other plans and metered API open as it adds compute.
- Tempus gets up to $9.5M to test Olivia as autonomous heart-care AI
Eric Lefkofsky's Tempus will develop and validate a heart-failure agent under an FDA-engaged federal program, though its authorization path and treatment authority remain unsettled.
- Feyn Labs launches MultiMatte to keep the object you name and erase the rest
Shreyash Nigam and Hafedh Hichri fine-tuned 2.27% of Meta's SAM 3, preserving text prompts while replacing binary masks with alpha mattes.
- Summation opens its AI analyst to self-serve teams at $60 per user
Opendoor co-founder Ian Wong is betting recurring, traceable analysis can move from enterprise deployments to everyday teams.
- Anthropic says Moonshot routed Kimi user requests through Claude
Anthropic alleges Moonshot sent nearly 300,000 Kimi customer requests primarily to Claude Opus through 5,380 fraudulent accounts.
- Epsilon Health raises $27.6M to run an AI radiology practice
Founder Rustin Rassoli is building a medical practice around its software, rather than selling a standalone imaging model to hospitals.
- Cognition ships SWE-2, a cheaper coding model for Devin
Cognition says SWE-2 approaches Fable 5.1's coding performance at 36% of its rollout cost, based on the company's own benchmark.
- Skild AI hits a $100M run rate 10 months into selling robot brains
Deepak Pathak's robotics software maker reached the figure after raising $1.4B at a valuation above $14B in January.
- Embedder plugs its firmware agent into SEGGER's hardware debugging stack
Embedder's YC-backed agent can flash, inspect and debug firmware through SEGGER probes, moving AI coding onto the physical bench.
- Neurophos hires a Kymeta founder to commercialize its 2028 photonic chip
The Austin chipmaker is turning January's $110M Series A into hires, facilities and a commercialization bench for a photonic processor due in 2028.