RuntimeWire

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

Latest stories — page 21

• BMW gives Brett Adcock the factory proof Figure AI needed

  BMW says Figure 02 supported production of more than 30,000 X3s, while Figure 03 moves next into Spartanburg logistics work.

• Wayve opens $85 million employee tender as Alex Kendall turns autonomy hype into a talent fight

  The secondary sale gives staff liquidity at an $8.5 billion valuation while Wayve pushes toward Uber pilots and Nissan integrations.

• Head to head: Bytedance Seedance V1.5 Pro Image To Video vs Wan v2.6 Image to Video

  This matchup wasn’t especially close. Across both tests, Bytedance Seedance V1.5 Pro Image To Video was the model that actually obeyed the shot brief, while Wan v2.6 Image to Video kept drifting toward attractive but less correct imagery.

• EquiLibre's DeepMind founders turn poker AI into a more than $500M quant-trading bet

  The Prague lab raised an undisclosed Series A led by Creandum after taking reinforcement learning from games into live markets.

• Commerce is expected to lift Anthropic Fable 5 export controls tonight

  The move would reopen access to Anthropic's general-use frontier model after an 18-day fight over AI, cybersecurity and export law.

• Peter Beck's $8B Iridium deal gives Rocket Lab the network it did not want to build from scratch

  Rocket Lab agreed to pay $54 a share for Iridium, adding L-band spectrum, 2.55 million subscribers, and recurring satellite revenue.

• Head to head: AuraFlow vs Fibo Bbq Preview

  This one wasn’t especially close: AuraFlow has style, but Fibo Bbq Preview is the model that more reliably obeys the brief when the prompt gets fussy. Across three very different image tasks, Fibo won on compositional discipline and object-level accuracy, while AuraFlow’s best showing came when motion and mood mattered

• Conception Says It Has Made Early Human Egg Cells From Stem Cells

  Matt Krisiloff's fertility startup still has to mature the cells and prove safety before any clinical use.

• Realta Fusion shows its mirror reactor can pull electricity straight from plasma

  Kieran Furlong's UW-Madison spinout says the June 19 WHAM demo produced current at about 100 volts, but not net electricity.

• Google brings Gemini voice search into Gmail beta

  The beta turns inbox retrieval into a paid Gemini workflow across Gmail, Docs and Keep, tightening Google's grip on AI productivity.

• Clicks shows its keyboard phone actually working

  Adrian Li Mow Ching and his co-founders are moving the $499 Communicator from CES concept toward a late-2026 shipment.

• Etched exits stealth with $800M raised and a broader inference bet

  The San Jose chip startup says it has working silicon, first racks built and more than $1B in signed customer contracts.

• Google ships Nano Banana 2 Lite and Gemini Omni Flash to developers

  The new Gemini media models put faster image generation and conversational video editing into AI Studio, the Gemini API and enterprise tooling.

• Anthropic launches Claude Sonnet 5 as developers wait on price, context and model docs

  Anthropic has now announced Claude Sonnet 5 through the official Claude account, shifting the story from leak-watch to whether the new Sonnet tier delivers the claimed price cut, 1M context and near-Opus positioning.

• Claude Code's Hidden China Signal

  A reverse-engineering post claims Anthropic's CLI mutates hidden prompt text when proxies, time zones, or China-related domains appear, raising a bigger question: what else is your coding agent saying before you ever type?

• DeepReinforce releases Ornith-1.0 for self-scaffolding coding agents

  The MIT-licensed model family spans 9B to 397B parameters, but its benchmark lead rests on DeepReinforce's own harness-heavy evaluations.

• Head to head: grok-4.3 vs Llama-3.3-70B-Instruct

  This matchup wasn’t close: grok-4.3 won on both instruction discipline and editorial reliability. Across code, JSON formatting, and business writing, it consistently did the exact job asked, while Llama-3.3-70B-Instruct kept slipping on output constraints and precision.

• Omnea founder Ben Freeman wants to turn employee churn into a $250K seed check

  The London procurement AI company will let five-year employees pitch for $250,000, even though no one is eligible yet.

• Robert Scoble is turning his AI-founder Rolodex into a paid intelligence product

  Scoble says his lists include 13,000 AI founders, but the more important bet is packaging a hand-curated X graph as startup discovery infrastructure.

• Kali Linux 2026.2 puts maintenance ahead of spectacle

  The new release adds 9 tools, faster VM boots, GNOME 50, KDE Plasma 6.6 and a heavier NetHunter push for Android-based testing.

• Meituan's LongCat team releases 1.6T-parameter AI model trained on domestic ASICs

  LongCat-2.0 is billed as an open-source MoE model, but the release still says full weights are coming soon.

• Flytrex and Wing's Dallas drone pact turns air traffic into the bottleneck

  The rivals say their shared UTM layer handled about 8,000 overlapping Dallas-area deliveries with zero airspace conflicts.

• Halvar Flake writes the startup guide venture firms will not

  The zynamics and optimyze founder published a 28-page guide that treats VC money as a tradeoff, not a milestone.

• Base44's founder takes the expensive route to defensibility: owning the model

  Maor Shlomo's Wix-owned app builder is rolling out Base 1, a custom LLM trained on real product-building sessions.