Frontier labs, model releases, capability trends and the economics of inference.
28 articles
The U.S. Department of War designated Anthropic a national security supply chain risk after negotiations over a $200M Pentagon contract collapsed. The dispute centered on Anthropic’s refusal to remove guardrails preventing its AI from being used for mass domestic surveillance or lethal autonomous we
Agents like OpenClaw may reduce hiring pressure by absorbing recurring operational work that historically justified incremental headcount. Instead of expanding payroll, teams can scale variable token spend, allowing smaller groups to operate with higher leverage while delaying permanent hiring commi
This little bugger just broke the internet. Who is he?
Internal AI agents and collapsing software development costs are eroding the traditional moat of enterprise SaaS as intelligence shifts inward. As externally packaged software becomes infrastructure rather than the primary source of differentiation, public markets are repricing incumbents around wea
OpenAI’s code red is less about falling behind technically and more about recognizing that momentum has shifted. As foundation models converge, advantage increasingly comes from distribution, trust, and integration rather than raw benchmark performance. Google and Anthropic are gaining ground not by
Day by day, it is becoming abundantly clear that the AI models powering Western companies are no longer the ones made in the West. That reality has slipped in almost unnoticed. While the public conversation circles around frontier labs and safety debates, engineers inside North American companies ar
Artificial intelligence has become astonishingly good at perception. It can describe a photo, summarize a document, and even generate a symphony of pixels from a sentence. The models behind these feats (convolutional networks, transformers, diffusion systems) have mastered how to recognize patterns.
Apple has shoveled $704B into buybacks over the last decade, clear signal it’s now an EPS-first incumbent, not a moonshot machine. “Apple Intelligence” is incremental, Siri 2.0 slipped to 2026, and Siri punts tough queries to ChatGPT while Google/Samsung push true AI-first phones. Cash returns run ~
GPT-5’s launch highlighted the gap between lofty expectations and the practical realities of AI progress. While bold claims about AGI have been tempered, the real advances may lie in efficiency, cost, and usability rather than sudden leaps in intelligence. Much like electricity or the internet, the
OpenAI’s GPT-5 is its most advanced model yet, using a new Mixture-of-Experts design for faster, more accurate, and more versatile performance across text, images, and audio. It offers deeper reasoning, stronger coding/writing, a massive 400K-token context window, and improved safety. New “Agent” to
Figma’s IPO filing included language authorizing blockchain-based common stock, signaling openness to tokenized equity. While they haven’t issued any yet, this move could lay the groundwork for more flexible liquidity options in private markets.
In June 2025, Meta acquired 49% of Scale AI, turning a neutral infrastructure provider into a strategic asset. OpenAI and Google cut ties in response. By installing Scale’s founder to lead its new Superintelligence unit, Meta is accelerating its push to close the AI innovation gap.
VE Weekly
One send a week. Whole pieces, not a digest of links out.