RuntimeWire

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

Latest stories — page 38

• Addy Osmani packages Claude Code agent skills into slash commands that mirror the dev cycle

  A post says the kit wraps senior-engineer patterns into slash commands like /spec, /plan, /build, /test, /review, /ship; it claims seven commands but lists six.

• Mystery company allegedly spent $500 million on Claude in one month after leaving license usage uncapped

  Tom's Hardware, citing Axios, says an unnamed enterprise forgot to set usage limits on employee Claude licenses, amplifying worries that corporate AI spend is outpacing returns.

• Helios runs LIDAR-based plug-in solar math for any British address

  The browser tool ray-traces LIDAR to estimate balcony solar yield and payback per address, reflecting UK rules like the 800 W cap and MCS export limits.

• Andrew Kelley rewires Zig build: new configurer-maker split aims for 90% faster runs

  The Zig creator says a reworked build pipeline that caches config and runs an optimized maker process is on main and slated for 0.17.0.

• Francis (@fdavidsont) is about to launch a travel tool, per Ritwik Pavan

  Ritwik Pavan says he saw an early build and called it an awesome tool for travelers, linking to @fdavidsont ahead of a Monday launch.

• Caelan Garrett to present ScheduleStream, a GPU-driven multi-arm planner, at ICRA 2026 in Vienna

  The researcher says ScheduleStream tackles multi-arm task-and-motion planning on GPUs; the announcement surfaced via an X post amplified by Bowen Li.

• Xiaoxuan Ma shares REST3D, aiming for physically stable, visually consistent 3D from a single photo

  In a post on X amplified by Bowen Li, the REST3D project teases single-image 3D scene reconstruction; no paper or code link was provided in the announcement.

• Liquid AI ships LFM2.5-8B-A1B, an 8B on-device MoE trained on 38T tokens

  The on-device 8B MoE adds a 128K context, 128K vocab, and scaled pretraining to improve tool-calling on laptops, with base and post-trained weights on Hugging Face.

• What accelerators take and give: 1752vc lists terms for 45+ programs

  The X thread links to a single 1752.vc page with application links; examples include Boost VC at $500K for 15% and Techstars at $220K for 5%+ via SAFE.

• FROST paper claims a web page can ID your other sites and apps via SSD activity

  A paper linked in a thread says FROST can identify websites with 88.95% accuracy and applications with 95.83% by measuring SSD activity from a visited page.

• Report: Asana is buying StackAI for $75M to anchor an Agent OS

  A paywalled report claims Asana is acquiring YC-backed StackAI for $75M to power an Agent OS; there is no public confirmation and key deal details remain unverified.

• Solo founder says YC accepted him to build his AI clone as a money-making operator

  Solo builder Human Thomas says he got into YC with $40k monthly revenue and is launching Thomas, a virtual human whose only job is to make money.

• NVIDIA releases NVFP4-quantized Qwen3.6-35B checkpoint on Hugging Face

  4-bit NVFP4 build of Alibaba’s Qwen3.6-35B-A3B lands with Apache-2.0 license, long-context support, and vLLM instructions; NVIDIA notes it did not develop the base model.

• Gradio's "Build Small" hackathon opens on Hugging Face with $40k+ in prizes; registration closes June 3

  The Hugging Face org for "Build Small" is live, with anchor sponsors OpenBMB, OpenAI, and NVIDIA; small-model constraints, two tracks, bonus-quest badges, and a two-weekend build window.

• Ollama adds OpenJarvis, a local-first personal AI from Stanford labs

  Built with Stanford's Hazy Research and Scaling Intelligence labs under their Intelligence Per Watt program, OpenJarvis now runs locally via Ollama.

• Figure runs $100M employee tender, Brett Adcock says

  Adcock framed the secondary as a retention and hiring tool as Figure pursues general robotics.

• Kog says it hit 3,000 tokens/s per request on standard GPUs

  Kog opened a live coding playground and argues single-request decode speed, not FLOPS, is the bottleneck that matters for autonomous agents.

• 120k users, 325k+ designs: YC touts Drafted, an AI home design launch from Nick Donahue

  YC says Drafted turns a hand-drawn layout and constraints into floor plans and 3D designs; the free app focuses on early-stage, single-level layouts. YC adds 120k users made 325k+ designs in the past month.

• Chad Whitacre says he is retiring from tech to live offline

  The open source builder disclosed the move in a brief post, writing that AI took the last of the wind out of his open source sails, and noting he works for Sentry.

• X user @sang_yun_lee tees up 'sleep' for language models in new work

  In a short X post, the user @sang_yun_lee hints at a periodic, recurrent cycle for LMs, framing the idea with: "Almost all animals sleep. Why don't LMs?"

• XCENA raises $135M to make memory, not compute, the center of AI performance

  CEO Jin Kim and a Samsung/SK hynix veteran team rebrand MetisX as XCENA and double down on CXL computational memory with MX1, at a $570M valuation.

• AISlop ships CLI and GitHub Action to catch AI-generated code smells

  AISlop is live on GitHub and the Marketplace as a CI quality gate with ai-slop/* rules, signaling a push to flag AI-written code in CI workflows.

• AI drug discovery enters physical validation: Insilico, Recursion, FutureHouse and Google DeepMind credited with new milestones

  Peer-reviewed data for AI-generated peptides, enzymes and CRISPR variants, plus AI-designed compounds entering human trials, push in silico models into preclinical reality.

• Brett Adcock says Figure AI deployments are imminent, staking humanoid race on intelligence

  The Figure AI CEO says multiple customer deployments will be announced within 90 days as Figure 03 shifts from demos to production work.