structure • logic • style

Joined December 2017
448 Photos and videos
Pinned Tweet
This is the most exciting miniseries I've ever watched. I am truly on the edge of my seat.
4
141
let the dreamers dream
x.com/vitrupo/status/1926074… That's a really interesting thought! Demis Hassabis says that hallucinations aren't necessarily bad. They helped AlphaEvolve develop the new algorithm. That's surprising! Should we rethink hallucinations? As creativity?
1
230
if you are on fire, do it on fire
24 May 2025
1
3
319
Excellent overview of AlphaEvolve new level unlock x.com/MLStreetTalk/status/19…

1
116
get your prompts ready
Our early work on core algorithm improvements using Gemini, called Alpha Evolve. Great to see this type of progress! deepmind.google/discover/blo…
1
132
"drastic agents" on the way
Ahead of I/O, we’re releasing an updated Gemini 2.5 Pro! It’s now #1 on WebDevArena leaderboard, breaking the 1400 ELO barrier! 🥇 Our most advanced coding model yet, with stronger performance on code transformation & editing. Excited to build drastic agents on top of this!
1
184
phi ARCHITECT retweeted
Excited to be full time on ARC-AGI-2. It's going to be a slow cook for the new ideas to be realized. Maybe we'll see a new face or two on the team! 🚀 @arcprize @MohamedOsmanML @bayesilicon @DriesSmit1
4
5
73
6,931
observation at the edge
1
5
140
phi ARCHITECT retweeted
This week I had Gemini assist me on switching from vim to neovim. I avoided it for years because of the burden it would be on me. I gave Gemini my current configs and started a deep research doc on making the change. It produced a 53 page book on everything I would need to know. I then used this info with 2.5 Pro in aider to implement all the recommendations. Nearly perfect from the start. Less than a day - from start to finish. Since then we've configured the debugger and a number of other fine-tunes. This would have taken 1-2 weeks for me to understand and evaluate all the options. My only question right now is what other tasks on my plate can I accelerate in this way. github.com/photon-platform/n…
1
12
960
reverse tsunami
1
2
262
pay ATTENTION:
As expected, that was popular. Here is my attempt at consolidating all the answers into a list. - Prenorm: normalization in the residual blocks before the attention operation and the FFN respectively - GQA (Group Query Attention): more Q than (K, V)
1
123
phi ARCHITECT retweeted
18 Apr 2025
🐰 Easter Surprise! From now until April 20th, Say, Pi has UNLIMITED TTS—no quotas, no limits, just voices! 🎙️ Perfect time to explore our ALTER voice mode for Claude AI. Hop to it! #SayPi #EasterSpecial #Unlimited 🥳 #ClaudeAI
2
2
2
457
it seems Claude has native support for rendering mermaid docs it handled SEER's workflow without much difficulty
1
3
127
just prompt: create an artifact for a mermaid diagram for the following scenario...
2
74
testing the ARColator
1
2
428
I was just given access to Veo2 Now I can't find my way out @GoogleLabs
2
4
22
4,322
phi ARCHITECT retweeted
15 Apr 2025
Breaking news, @saypi_ai fans! 🚀 #SayPi is now available for #ClaudeAI by @AnthropicAI! 🧡 Experience the magic of voice interaction today on claude.ai! 🗣️ 💬 🤖 Meet #ALTER, the unique voice personally trained by @phi_architect! Phi has been working closely with us to finally give #Claude a voice, creating an unforgettable chat experience using our independently developed browser extension! 🚀 We're incredibly grateful to Phi for this fantastic collaboration! 🥳 Read the full announcement in our blog here: saypi.ai/blog/announcing-say… If you're already using Say, Pi, you can call claude.ai right away after reactivating the browser extension! 📞 🤩 Available for Chrome, Firefox and Edge. Safari coming soon. Hey @Pi friends, don't worry! Our beloved #PiAI by @inflectionAI will definitely remain a lovely friend to us, and we're still committed to supporting and developing for it further! 💚 🥧
2
3
5
390
SEER has been doing a lot of thinking
1
12
589
soon we will be conversing with dolphins, only to find out how truly stupid and misguided we are
DolphinGemma is a foundational AI model trained to learn the structure of dolphin vocalizations and generate novel dolphin-like sound sequences. 🐬
1
199
"...the processes that lead to (human) cognition are just mechanisms - so why couldn't you put them into a machine and have the machine do what we do?" ~ Mazviita Chirimuuta
🧠 Why does AI *seem* so inevitable to us? 🧠 Prof. Mazviita Chirimuuta, author of the Brain Abstracted explains the shifts in thinking over many decades towards a mechanistic understanding of intelligence.
1
2
151