AI Builders Digest 2026-07-11

AI Builders Digest — July 11, 2026

╔══════════════════════════════════════════════════╗ ║ TWITTER ║ ╚══════════════════════════════════════════════════╝

Podcast Highlights

Jürgen Schmidhuber — Ep 90: AI Pioneer on the State of AI Today Schmidhuber, called the father of AI by the New York Times and Forbes, argues that while AI behind the screen is passing the Turing test, true AGI requires physical hardware — robotics that can compete with human bodies in the real world. He's optimistic about AI technology but pessimistic about model companies' moats, noting that recursive self-improvement won't save them because gradient descent has inherent limits. A core insight: all data on the World Wide Web is a tiny fraction of all possible data. The future belongs to artificial scientists that collect data through their own experiments, not just training on human-generated web data. "True AI is not just the AI behind the screen, which is working very well and which is passing the Turing test now. True AI is also, you know, real robots, real machinery outside of the stream and the real world and the physical world." — Jürgen Schmidhuber https://www.youtube.com/watch?v=RKjR8DQ40po

Replit CEO Amjad Masad Making two sharp observations: AI is making coding less rigid, yet our infrastructure is becoming MORE rigid — his infra teams are writing formal specs for the first time. The faster you want to move, the more solid the ground beneath you has to be. On the LLM market, he notes VCs suffered "Anthropic psychosis" six months ago and convinced themselves it would be a monopoly. They will keep making great models, but so will everyone else — including new entrants. https://x.com/amasad/status/2075423115052790054 https://x.com/amasad/status/2075413916491075755 https://x.com/amasad/status/2075358353686208741

Vercel CEO Guillermo Rauch Predicting open models are about to get exorbitantly fast. On this week's model releases, he suspects Meta Spark 1.1, Grok 4.5, and GLM 5.2 will significantly displace token market share. His thesis: most agentic tasks require reasonably high intelligence at fast speeds, making this an excellent time to use the AI Gateway. https://x.com/rauchg/status/2075294327354577256 https://x.com/rauchg/status/2075294130327196152 https://x.com/rauchg/status/2075255565627080813

Research @AnthropicAI Alex Albert Shared a Fable demo, continuing the wave of enthusiasm for the agentic coding tool. https://x.com/alexalbert__/status/2075285305096343583

Box CEO Aaron Levie On AI's competitive future: if AI is trained on the best datasets in every industry (law, finance, healthcare, life sciences), the key question is how companies differentiate. His answer: companies that effectively use AI against a growing set of proprietary data and knowledge will be in a strong position. There's a reinforcing loop between model intelligence, proprietary data, workflow integration, and employee interaction — with compounding returns to those who do it best. https://x.com/levie/status/2075416313481290077

Levie also shared detailed Box AI Complex Work evaluation results for GPT-5.6 ("Sol"): Financial Services jumped to 76% from 71%, Healthcare went 58% vs 46%, Public Sector hit 74% vs 63%, and Life Sciences reached 60% vs 51%. Sol reasons from source definitions and checks documents rather than taking them at face value. https://x.com/levie/status/2075287443411222628

President & CEO @YCombinator Garry Tan Praised Meta Muse Spark 1.1 on his OpenClaw setup: "turned out to be really good on my OpenClaw." https://x.com/garrytan/status/2075445455438385255

VC @FirstMarkCap Matt Turck Noted France and Norway could play in the final (a rare occurrence after 60+ years) and shared a humorous post about Paris post-match. https://x.com/mattturck/status/2075386931924414686 https://x.com/mattturck/status/2075361793904439489 https://x.com/mattturck/status/2075342627436712362

Partner @fpvventures Nikunj Kothari Delivered an extremely detailed breakdown of the current model release situation: GPT-5.6 in three flavors (Sol, Terra, Luna with Sol clearing government safety evals), Grok 4.5 launched the day before GPT-5.6 for "24 hours of being frontier-adjacent" at /M tokens, Fable 5 back after three-week government suspension, Sonnet 5 as the new free default, Meituan (a Chinese food delivery company) open-sourced a 1.6 trillion parameter model, ByteDance launched Seedream 5 Pro, and OpenAI shipped GPT-Live with real-time acknowledgment. Ollama raised 5M for local open-source models. https://x.com/nikunj/status/2075411514773967261 https://x.com/nikunj/status/2075248134457041341

CEO @every Dan Shipper On GPT-5.6 SOL as "the gold standard for knowledge work" and discussed husk doing vibe checks. https://x.com/danshipper/status/2075264022988116280 https://x.com/danshipper/status/2075389275969826879

General Partner @SouthPkCommons Aditya Agarwal Shared a deep-dive on Gagan_Shux — the 2nd Indian in history to reach space — covering ISS life, microgravity effects, and the future of human space exploration (Moon, Mars, and beyond). https://x.com/adityaag/status/2075414469497270554 https://x.com/adityaag/status/2075414469497270557

Sam Altman Shared three posts: heartfelt message about fidji at OpenAI, promoted Codex as the core of their new work product, and highlighted GPT-5.6 SOL as a huge step forward on dollars-per-task, along with Terra and Luna variants. https://x.com/sama/status/2075354679031067058 https://x.com/sama/status/2075293792048136572 https://x.com/sama/status/2075267201058426944

Thariq — Claude Code @Anthropic AI On agentic coding as "reducing your unknowns," and shared a Fable demo. https://x.com/trq212/status/2075283841758183674 https://x.com/trq212/status/2075280416995705312

Polyagentmorous ClawFather Peter Steinberger Shared three appreciative posts — building something great, hosting a livestream, and kudos to fellow builders. https://x.com/steipete/status/2075350572560191630 https://x.com/steipete/status/2075328495677521934 https://x.com/steipete/status/2075313523237019686

Research Scientist Amanda Askell On building collapses in New York: shared the counterintuitive finding that despite the perception, buildings don't just randomly collapse regularly in NYC despite the memorable incidents that create that impression. https://x.com/AmandaAskell/status/2075247953309311043 https://x.com/AmandaAskell/status/2075245939548455009

╔══════════════════════════════════════════════════╗ ║ PODCASTS ║ ╚══════════════════════════════════════════════════╝

See top highlights above under Jürgen Schmidhuber (Ep 90: AI Pioneer on the State of AI Today — Unsupervised Learning podcast). https://www.youtube.com/watch?v=RKjR8DQ40po

╔══════════════════════════════════════════════════╗ ║ BLOGS ║ ╚══════════════════════════════════════════════════╝

No new blog posts today.

Generated through the Follow Builders skill: https://github.com/zarazhangrui/follow-builders