Distillation only gets you as far as the teacher. You inherit the teacher's blind spots, and you cannot verify what it taught you. Parallax does not distill. We build a graph database of real, grounded sources and derive each question from its answer, verifying every…

Harvard × Chutes research data collection wrapping up The data-collection phase of our research collaboration with Harvard is coming to a close. The opt-in research endpoint (research-data-opt-in.chutes.…) will be removed this Friday, 24th July. If you're still routing…

"How much power and money can we sacrifice at the altar of AI to build God?" That is @jon_durbin, and that is only the first line of the trailer. It closes on the line we cannot wait to defend: "We proved that it's plausible. What we need to prove now is that it's not only…

Week 10. The rundown: → Jon pre-trained a 20B MoE for under $10/hour of compute. Eight rented L40S VMs across two continents plus a few consumer RTX cards. A public demo, not a benchmark. It happened → Revenue per trillion tokens hit $340K, a new all-time high. 90-day…

Article Last Week in Chutes | 14th July to 21st July Last Week in Chutes July 14 to July 21, 2026 Week ten. Jon pre-trained a 20B model for under $10 an hour of compute, in public, on rented VMs. Also this week: monetization hit an all-time high, Chutes