What used to take a 9-month build cycle now ships in 3 days with Replit. pic.twitter.com/yfeioc6dab
— Replit ⠕ (@Replit) 22 avril 2026
What used to take a 9-month build cycle now ships in 3 days with Replit.
AI
-
Replit Cuts 9-Month Build Cycle Down to 3 Days
By
–
-
New benchmark PerfectSquashBench for image model anchoring in AI
By
–
Image models tend to get much more stuck on a particular direction than text models, requiring clearing the context window fairly often. PerfectSquashBench is my new measure of how image models anchor. The squash remains merely fine after many attempts.
-
Odyssey-2 Max: Largest World Model with Real-Time Physics
By
–
🌍 Odyssey just launched Odyssey-2 Max — their largest world model yet
— Futurepedia – Learn to Leverage AI (@futurepedia_io) 22 avril 2026
It simulates physics in real time: gravity, collisions, fluids, biomechanics — all running live as you interact with it.
Scores highest in physical accuracy among all evaluated world models, with 3x the… pic.twitter.com/OiW66rgzysOdyssey just launched Odyssey-2 Max — their largest world model yet It simulates physics in real time: gravity, collisions, fluids, biomechanics — all running live as you interact with it. Scores highest in physical accuracy among all evaluated world models, with 3x the
-
Megawatt AI Infrastructure Demands 150,000 Electricians Urgently
By
–
Reminder: we are moving to one megawatt racks in Intelligence Factories.
— Nina Schick (@NinaDSchick) 22 avril 2026
We need more energy, a lot of it.
We also need a lot more people.
Like electricians: 150,000 are mission critical just to serve the ~90GW of Intelligence coming online in the next two years. https://t.co/5AK8AmtGWPReminder: we are moving to one megawatt racks in Intelligence Factories. We need more energy, a lot of it. We also need a lot more people. Like electricians: 150,000 are mission critical just to serve the ~90GW of Intelligence coming online in the next two years.
-
API Design Change: Rejecting Set Calls Without Force Parameter
By
–
Did some digging and see the confusion now. This is as designed, but I see how that seems confusing. Will change code so set calls are rejected unless you force, so users are steered to add calls.
-
Generating realistic images with AI in seconds
By
–
Exactement on peut maintenant totalement inventé des screen ultra réalistes en 2sc 🙂
-
Creating AI Agent Spaces Harder Than Model Training
By
–
the struggle is real haha sounds harder for the agent to create a space than to train a model
-
Recurring AI incidents tracked over thousand occurrences
By
–
This has in fact (at various levels) happened more than 1000 times. Check out @DamienCharlotin
’s tracker and my substack essays about it. -
Why Open-Weight Escape Hatches Matter for Embedding Models
By
–
This is why I won't use proprietary hosted embedding models myself – I am more than happy to pay for a hosted solution (cheaper, faster and more convenient than self-hosting) but I want an open weight escape hatch for if they ever stop serving it