# RuntimeWire – Startup intelligence for the AI economy > Source: > Published: 2026-07-16 03:16:46+00:00 Real-time startup intelligence for the AI economy. Funding, model launches, infra shifts, and founder moves — the signals that matter. [MIT researchers publish non-generative test for CSAM-tuned AI models](/article/mit-gaussian-probing-csam-ai-model-audit)Gaussian probing checks LoRA-adapted diffusion models by reading internal activations instead of producing illegal outputs. [Head to head: AnimateDiff vs Luma Ray 3.2 Image to Video](/article/head-to-head-animatediff-vs-luma-ray-3-2-image-to-video)One model showed up as a general-purpose image-to-video system; the other mostly showed isolated flashes of competence. Across four prompt types, the result wasn’t subtle. [EdVisorly raises $13.3 million to automate college admissions back offices](/article/edvisorly-raises-13-3-million-ai-college-admissions)Founder Manny Smith is using an Air Force systems background to sell EddyAI into transcript and transfer-credit workflows at more than 100 institutions. [Nous Research nears $75 million round at $1.5 billion valuation](/article/nous-research-75-million-round-1-5-billion-valuation)Robot Ventures is leading the reported financing, with USV participating, as the 2023 open-source AI lab pushes Hermes toward a business. [Whatnot acquires Shaped to speed up live-shopping recommendations](/article/whatnot-acquires-shaped-live-shopping-recommendations)Shaped founder Tullie Murrell and nearly a dozen engineers and AI researchers will join Whatnot as the $11.5B marketplace expands beyond collectibles. [Head to head: AuraFlow vs Imagineart 2.0 Preview](/article/head-to-head-auraflow-vs-imagineart-2-0-preview)One model flirted with flashes of taste; the other actually delivered on prompts. Across eight image tests, Imagineart 2.0 Preview separates itself with stronger prompt adherence, better spatial logic, and a statistically decisive win. [DeepSeek's V4-Pro price cut exposes the agent margin problem](/article/deepseek-v4-pro-price-cut-agent-margin-problem)Liang Wenfeng's lab cut API prices, but agent workflows can still turn one user request into dozens of billable model calls. [Saturn Cloud adds Lilac's idle GPU network for per-token inference](/article/saturn-cloud-lilac-idle-gpu-inference-partnership)Founder Sebastian Metti is betting enterprise-owned GPUs can feed Saturn's token factory without reserved capacity. [Thinking Machines releases Inkling, its first open-weights AI model](/article/thinking-machines-releases-inkling-open-weights-model)Mira Murati's lab is putting full weights on Hugging Face and routing customization through Tinker, its managed fine-tuning platform. [OpenAI ships GPT-Red to train GPT-5.6 against prompt injection attacks](/article/openai-gpt-red-automated-red-teamer-prompt-injection)The internal-only model uses adversarial self-play to generate attacks that harden OpenAI's production models before deployment. [Arbital Health launches Arbital Flex, a self-serve actuarial AI for value-based care contracts](/article/arbital-health-launches-flex-actuarial-ai)Brian Overstreet and Travis May are packaging Arbital's risk-contract engine for buyers that lack large actuarial teams. [Hacked Suno code shows YouTube and Deezer scraping behind AI music models](/article/suno-hacked-code-youtube-deezer-genius-scraping)404 Media's report lands weeks after Suno said it raised more than $400 million at a $5.4 billion valuation. [Head to head: Kimi-K2.7-Code vs gpt-5.4-mini](/article/head-to-head-kimi-k2-7-code-vs-gpt-5-4-mini)This was a close matchup on aggregate, but Kimi-K2.7-Code finished ahead by being more reliable on instruction-following and structured tasks. gpt-5.4-mini had real strengths in polished prose and one coding task, yet Kimi took more categories and the overall edge. [OpenBox AI integrates runtime governance with Temporal for long-running agents](/article/openbox-ai-temporal-runtime-governance-agents)The July 13 integration gives Temporal developers policy checks, approvals, attestations, and audit trails inside agent workflows. [Refer raises $7.5 million for candidate-paid AI recruiting](/article/refer-raises-7-5-million-candidate-paid-ai-recruiting)Founder Andre Hamra built Refer after putting up flyers at Stanford to help people find jobs; hires pay 20% of their first month's salary. [Float raises EUR4.5M to build AI finance stack for European startups](/article/float-raises-eur45m-ai-finance-stack-european-startups)CHAPTERS Group led the Series A and its CEO Jan-Hendrik Mohr will join the board as Float moves beyond lending. [DTU and ORCA validate quantum-AI peptide design in lab tests](/article/dtu-orca-quantum-ai-peptide-design-lab-tests)A July 9 preprint says a photonic quantum processor helped a generative model find MHC-binding peptides where data was scarce. [tinygrad maps GPU BARs in userspace for NVIDIA and AMD GPUs](/article/tinygrad-gpu-bars-userspace-nvidia-amd)George Hotz's AI stack still relies on Linux plumbing, but the code shows a serious attempt to move GPU control into a smaller runtime. [Stripe and Advent bid more than $53 billion for PayPal, Reuters says](/article/stripe-advent-paypal-53-billion-offer)A reported PayPal offer would move Patrick and John Collison's Stripe from payments infrastructure into large-scale consolidation. [Attovia Therapeutics files for Nasdaq IPO to fund immune-disease biologics](/article/attovia-therapeutics-files-nasdaq-ipo-atto)Tao Fu's San Carlos biotech is seeking an ATTO listing before disclosing price range, share count or valuation. [Head to head: AnimateDiff vs Marey Realism V1.5](/article/head-to-head-animatediff-vs-marey-realism-v1-5)This matchup wasn’t subtle: Marey Realism V1.5 swept all four tasks and did it with a decisive statistical margin. AnimateDiff had flashes of visual appeal, but across prompt adherence, motion, and scene construction, Marey was the model that more reliably delivered the shot that was actually asked for. [Tinygrad's founder rejects AI doom as a valuation pitch](/article/tinygrad-george-hotz-ai-hype-frontier-lab-valuations)George Hotz says he loves LLMs and coding agents, but argues frontier labs are selling fear to defend value capture. [Philips Hue ships Bridge Pro fix after firmware bricked some hubs](/article/philips-hue-bridge-pro-firmware-fix-unrecoverable-state)Firmware 2071401010 patches a failure mode that could leave the $99 Hue hub unrecoverable after a manual update. [LALAL.AI launches Lynx, 6x smaller model now default in Voice Cleaner](/article/lalal-ai-lynx-voice-isolation-model)The Swiss audio startup says Lynx is 6x smaller than Andromeda and is now the default model in Voice Cleaner.