my AI predictions for 2025: - at least one lab will declare AGI and mentions ASI - Q1: Google, Anthropic, OpenAI, META, Qwen and Mistral model fiesta ( it will be heaven ) - agents / computer use takes off - release of Claude 4, Gemini 3, GPT-5, Grok 4 (or whatever they call

I can't wait for Groq to host Kimi-K2 Thinking at 500 tokens/s

Ready to make your mark? At AEG, we turn ambition into action. Let’s bring your dreams to life. Join us

they should add some gardens and trees

I need to know the time horizon of Kimi-K2-Thinking on METR asap

I couldn't care less about GPT-4o just release GPT-6 already

The french government created an LLM leaderboard akin to lmarena, but rigged it so that Mistral Medium 3.1 would be at the top Mistral 3.1 Medium > Claude 4.5 Sonnet or Gemma3-4B and a bunch of Mistral models > GPT-5 ??????????????????? LMAO

One of the first women to become a physician in Britain was Elizabeth Garrett Anderson (1836-1917). She received education on the wards as a nurse, and then established herself on the medical register through the Society of Apothecaries. She founded the London School of Medicine

before the OpenRouter bois ask: it breaks over the API and it breaks in the OpenRouter Chat and yes with the API I've set the request timeout to like 1 hour

"the cost per unit of a given level of intelligence has fallen steeply; 40x per year is a reasonable estimate over the last few years" "In 2026, we expect AI to be capable of making very small discoveries. In 2028 and beyond, we are pretty confident we will have systems that can

OpenAI is pushing the narrative further towards "adults should be able to use AI on their own terms" (they will ship goonbots on christmas)

https://t.co/SBdSNcybY9

Kimi-K2 is a great uncensored model

Preached last week on "the church that ministers" and realized something: we've trained people to be efficient workers for God, but forgotten to teach them to be worshipers. Adoration isn't productivity. It's the restful acknowledgment that God doesn't need our resume - He

the french are completely losing it (me too, haven't stopped laughing in like 3 mins, this is hysterical)

https://t.co/prZLEF80PV

The french government created an LLM leaderboard akin to lmarena, but rigged it so that Mistral Medium 3.1 would be at the top Mistral 3.1 Medium > Claude 4.5 Sonnet or Gemma3-4B and a bunch of Mistral models > GPT-5 ??????????????????? LMAO

Kimi-K2 Thinking crushes all frontier models except GPT-5-high on the Artificial Intelligence Index Open-Source models closed the gap!

Anthropic continues going vertical > September: $170B > Now: "Anthropic could target a valuation between $300 billion and $400 billion"

I thought using the Turbo endpoint would resolve the issue, but nope. Only made it 3 times more expensive

2 years ago veteran endurance athlete, Kurt Madden, was given a few pouches of VESPA. Initially he blew it off, but due to the persistence of his friend Mike McKovitch, he decided to give it a whirl on a training brick (bike/run) and his world changed . . . . . . fast forward to

so should only be about $2

actually only ~255k output tokens

Kimi API is much faster now but still cooked. A lot of long requests just break after a few minutes of reasoning. You get charged but have no results. > was happy that I finally got some requests through for LisanBench with Kimi-K2-Thinking Turbo > used ~300k tokens@$8/million

GPT-4, Claude-3 Opus and GPT-4.5 are a gift to the world

AI agents at enterprise scale ⚡️ Designing the Giga brand. With the brilliant @varunvummadi @eshamanideep @jcarvajalpa