Pinned Tweet
Day 4 of memefying myself this time for my product hunt launch. producthunt.com/posts/photog……
#buildinginpublic @_buildspace #indiehackers @_nightsweekends
Claude instantly guesses GPT's word and GPT wins??
We gave 4 AI models a secret word each and told them to steal the others.
GPT, Claude, Grok and DeepSeek bluffed and interrogated each other while an AI judge scored the smartest player. GPT won by barely speaking: it planted one idea early and let the rest expose themselves.
All four ran on one AI/ML API key. Code and full transcript below:
52
I don't know anything about this but the aesthetics are on point 😍
Blume.codes 1.0.47 is live! Better agentic codebases🌸 Blume reads your agent conversations locally, then promotes steering and pain signals into code to make the next run better and cheaper.
Stop repeating yourself to coding agents - Live on Mac, Linux and Windows
1
1
41
The search queries that @perplexity_ai performs and the the results it gets are a serious bottleneck (@MrBeast )
This is the opposite of grounding as the LLM's are
1. Gaslit into thinking the query aligned with the request
2. Forced to consolidate the results or ignore them
2
1
170
Big if true
Introducing SubQ - a major breakthrough in LLM intelligence.
It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA),
And the first frontier model with a 12 million token context window which is:
- 52x faster than FlashAttention at 1MM tokens
- Less than 5% the cost of Opus
Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention).
Only a small fraction actually matter.
@subquadratic finds and focuses only on the ones that do.
That's nearly 1,000x less compute and a new way for LLMs to scale.
1
28
Hey @perplexity_ai why is half the screen empty and the settings part of the chat obscured by the not closeable ad? What's even better clicking the ad by accident deletes what you've been writing thus far 😐
31
I asked Claude for a plan and generated a 16 page word.docx - impressive. But very much overkill for what I expected to get.
Would be nice to get asked, before Claude starts writing code for 3min to generate a word doc.
@AnthropicAI
1
74
Whatever startup has time for that is ngmi
I collaborated with @wabi and @joonasvirtanen to bring this incredible intro sequence to life, and I’d love to share it with you!
We wanted way more flexibility than Liquid Glass, so I made a custom glass shader that refracts + disperses light.
More technical details soon :)
66
Hey @fal it's been like this for a couple days 😭
The profile page keeps loading with an 429 error.
(Too Many Requests)
Chrome with no extensions.
1
246
Thanks for the debugging @1littlecoder all fixed now. Well done! 🎖️
1
1
51
I pronounce @vite_js like kite with a v, because otherwise the STT in vscode spells it veet.
I think that settles the pronunciation debate.
1
1
157
Might be the first time where I don't understand the explanation of how a technology works 😅
crazy paper that combines high efficiency volume Bragg grating (VBG), spatial light modulator (SLM) for phase pattern, proper balance of coherence for light source, MEMS steering for color-angle multiplexing, and a neural waveguide model i don't understand yet, altogether to form a holographic MR display... won't be able to afford it for sure.
oh there is also a multilayer "apochromatic holographic eyepiece lens" that cancels the chromatic dispersion for the RGB color..
(from Stanford and Meta Reality Labs)
60
literally on every flight on every date right at the start.. 🤦♂️
@wizzair
2
104























