I like Paper Clubbing | AI Transformation lead at Formidium ($34B AuA) Prev. partner at @elixir_capital & @Woodstockfund | Electrical Engineer @bitspilaniindia
Joined October 2020
- Tweets429
- Following1.4K
- Followers63
- Likes1.1K
Pinned Tweet
How do you feel?
It's the oldest question in art and the newest one we can answer with technology.
MindReader takes any piece of content - a video, a post, a sales call - and simulates, second by second, how a brain responds. Where attention holds. Where effort spikes. What actually gets encoded into memory.
I've been running it with sales teams who want to know how a pitch landed in the buyer's head - deeper than the transcript, to see where attention dropped and where something stuck. We finally have* evals for conversation.
MindReader is built on Meta FAIR's TRIBE v2 + 35yrs of neuro research. The makers are engineers who've authored multiple academic papers and have been building open source for the last 4+ years.
Completely open source - I encourage you to tinker!
Run your latest social media post to see where attention dropped (or explore brain effort, gut reaction and 4 other neuro-dimensions).
The Product Hunt link is below
From Hierarchy to Intelligence
I was sceptical after reading it. The premise is certainly has hints of sci-fi*
Then I looked up the authors. Jack Dorsey and Roelof B. Truly fascinating read
How Block is using AI to eliminate hierarchical bottlenecks, building the first company organized as intelligence rather than hierarchy.block.xyz/inside/from-hierar… via @blocks
After months of manually and religiously seeding my new @X account well, I have a good filter:
(followers i know)/followers ratio
this profile-- it belongs, to a higher circle
Smile and wave, just smile and wave B)
BESTEST gender ratio && kickass community genesis && builder resources abundance && GREAT kebabs
Also ty @clairevo - you are a @bot template wizard 🪄
Notes on AI research papers:
1. Toward Generalist Autonomous Research via Hypothesis-Tree Refinement @namespace_ERI
@bigbean777
I was deeply excited to read this, given my background in research. The paper treats research memory as a data structure, not a longer transcript. one coordinator holds the map; short-lived executors test one branch each in isolated worktrees. every failed experiment stays attached to the hypothesis that produced it - evidence packets ftwwww.
It is truly every researcher's dream to have an intern who is so ready to work.
We discussed id the held-out merge gate can leak back to the HTR after, say 20+ merges. Another major drawback was that the HTR, given how trees work, was unable to propose drastically different solutions when things aren't working.
Brute force intern. Needs fast feedback. Read with 90/30 paper club
ok more notes on paper 4 later but heree is paper 5 - arxiv.org/pdf/2606.05661
CONTINUAL LEARNING BENCH: Evaluating Frontier AI Systems in Real-World Stateful Environments
The authors create a benchmark that separates being intelligent already from actually becoming better through experience, and find that today’s AI agents are surprisingly poor at the second one.
I wanna know if someone in the eval’s community has an answer for this: when it comes to cloning voice, how do you protect celebrity voices or podcast the voices? If you assume it could be through content (eg David from @AcquiredFM has a v signature intro), but how do you differentiate my very enthusiastic friend from MrBeast ?
Running a check of all voice samples - all phrases that MrBeast said would you be too computationally expensive - esp every time someone wants to clone a voice, especially in the free accounts. what is the safety eval?