AI coding is getting faster but the infrastructure behind it needs to catch up. slow CI means more idle time, more waiting and less time actually shipping. owen has a great breakdown of how Boardy cut CI from 40 minutes to just 2. runner wait: 11 min → 8 sec lint: 2 min → 5 sec typecheck: 70 sec → under 5 cache misses: 50% → 1% solid engineering work. this is a bottleneck more teams need to tackle.

Opus 5.5 turning NVIDIA simready models into a live warehouse is wild. curious how much steering it actually needed.

Anthropic vs OpenAI so far in 2026 - Clearly OAI got back in the game with steady gains in spend share throughout the year - Tokens go kinda wild with the introduction of Luna + new Haiku - Opus 5.5 reversed the trend Sweet, sweet competition

Which startups are most likely to get an investment from top tier VCs? I looked at ~60 attributes across 5000 startups and these were the 15 strongest. #1 is being from a top tier fund. MBA is a negative signal Full list of signals below:

no one can beat Google on this benchmark.

google got its TPU prototype to process a prompt in orbit and send a response back to earth. the real test is handling radiation and extreme temperatures. i think space based compute could open up a completely new direction for AI

Two banked Codex resets? Feels like OpenAI is saving something big for the finale.

Holy shit no more 5 hour limits on Claude Ultra?? this is so damn cool

Google is pushing AI infrastructure beyond Earth. - First AI inference in orbit - TPU processed its first prompt - Complete response sent to Earth - Project Suncatcher moves beyond data centers AI, quantum computing, Waymo and Gemini. Google is becoming a serious rival to Nvidia

artificial analysis just updated its model rankings and the new numbers are interesting. v5 now uses private test sets or held-out answers for 60% of its evaluations. • claude opus 5.5 leads at 58, followed by gpt-6 astra at 57. • the open weights gap widened from 3 points to 12. • model costs vary by over 13x per task. better benchmarks make it easier to pick the right model for the job, instead of just trusting the hype.

I usually look at model performance first but cost per task tells the real story.

Last night an AI agent performed brain surgery on a mouse. We gave it 19 gene therapy vectors in a spreadsheet, a mouse model of epilepsy, access to a motorize surgery rig with cameras and one goal: reduce seizures. It picked the drug. It picked the brain region. It derived the coordinates from three papers. Then it drove the stereotaxic arms into the brain and injected 250 nanolitres while watching the entry site through the rig camera for bleeding. A licensed vet gated every irreversible step and held a physical kill switch to stop the AI agent. Thread on what it decided, what it saw, and the moment it refused to let me override it. [Writing this on the spot out of excitement, but I've been 18hrs in front of my laptop coding and coordinating this, so keeping it brief and descriptive, more cohesive thoughts on this soon...]

Made myself an interaction lab. It’s a system that generates new interactions for me on a daily basis based on some predefined rules. It’s all pretty rough, but a good source of inspiration. I only keep the ones I like, so it reflects my preferences more and more over time.

two commands, one creator: monid run tikhub video_search "home coffee" > 30 videos, $0.001. monid run tikhub user_posts @funlife4us > median 1,011, $0.001 3,843,744 ÷ 1,011 = 3,802× their own baseline. $0.003 total. @typesafeai Jev, @MonidHQ, provider list price.

i keep looking at this chart because the gap is so big same 47 real tasks on live sites, median time from request to answer. with Sonnet 5, AgentWeb is 13x faster than browser use. the model doesn’t have to click through the website step by step anymore. $0.002 per action

Artificial Analysis v4.3.2 to v5, AI models re-ranking step by step.

ok claude solve this case, make no mistakes.

nanoGPT going from 39.9s to 6.7s is ridiculous. herman just pushed the record to a whole new level. - 83.2% less time - 500x faster than python - 11x faster than the best auto-researched rust code for me this is proof that human engineering skills still go a long way, especially with ai in the mix.

google dropped argon just 2 weeks ago and now carbon is #1. they are moving FAST