danny gm @danmodels0

A mobile app for simple personal health context: https://t.co/oej7IVZubk yokohama Joined March 2025

Tweets

101
Followers

24
Following

54
Likes

320

danny gm @danmodels0

3 days ago

One important thing to get right by AI developers: finding what layers/components really need HITL, considering model improvement timeline. Commiting to the wrong layer can be costly.

0 0 1 6 0

View Details

danny gm @danmodels0

4 days ago

How about just allowing users to input free text to describe preferences directly

Two important changes to timeline recommendations: - New AI/SI ranking model that predicts probable time spent much better - Retrieval model trained to surface posts people read all the way through User active seconds is growing exponentially and hitting all-time highs almost

2K 2K 19K 3.1M 2K

0 0 0 3 0

View Details

danny gm @danmodels0

6 days ago

AI labs probably training models exactly like this. Find a target (anything beautiful made by humans, software, movies, games, books...), REA, hill climb, repeat.

Computer @AskPerplexity

a week ago

REA (Reverse Engineer Anything) by @morluto is now a skill in Computer. Ask Computer to reverse-engineer an app and it inspects how the app works without its source code, then builds a working version. Tasks that use /reverse-engineer are free through October 31, up to 10,000

99 365 4K 910K 6K

0 0 0 18 0

View Details

danny gm @danmodels0

6 days ago

Btw, X (and Youtube etc.) should have a feature for free text input to directly describe kinds of content you want/don't want on recommended feed.

Elon Musk @elonmusk

a week ago

Please report ugly or misleading ads to 𝕏 @ads

3K 6K 47K 5.9M 2K

0 0 0 3 0

View Details

Lionel Mora @lionel_mora

a week ago

In July we open-sourced a dynamic graph of everything a child learns in primary school. 1,590 concepts, 3,221 dependencies. Now we’re turning every concept into interactive experiments kids can play and learn from. First one, age 7: heat and water. Everything that used to be a boring textbook can now be an engaging experience, potentially tailored to every child. Kids education just got a massive upgrade.

208 839 8K 433K 10K

View Details

danny gm @danmodels0

7 days ago

@lionel_mora Graph as UI can be tricky. But this genuinely looks like a good idea overall.

0 0 1 118 0

View Details

danny gm @danmodels0

2 weeks ago

@aaryankushwah Useful!

0 0 0 34 0

View Details

danny gm @danmodels0

2 weeks ago

@hardmaru Some nice open-ended / no-verifier properties it seems: - tasks are no so verifiable and long-horizon - whether a trace is “successful” is selected by relative self-evaluation.

0 0 0 66 0

View Details

danny gm @danmodels0

2 weeks ago

Publishing papers with gatekeeping is needed for easier communication given limited human cognitive capacity. Result is binary outcomes with bias without nuance. With AI (expanding cognitive capacity) do you still need it?

0 0 0 8 0

View Details

danny gm @danmodels0

2 weeks ago

@grok @0xpasan okay (disliked)

0 0 0 3 0

View Details

danny gm @danmodels0

2 weeks ago

@bryan_johnson 1. perfect guardrails for major humanity ending risks? 2. cure my dog health issues 3. remove bomb planted inside... sorry wrong thread

1 0 0 38 0

View Details

danny gm @danmodels0

2 weeks ago

@levie So primitives here are: orchestrating agent -> tasks/subagents/workflows + agent to agent messaging mechanism. Wonder how structure would evolve (to handle general tasks).

0 0 0 399 0

View Details

danny gm @danmodels0

2 weeks ago

@PrimeIntellect @grok How does agent-to-agent messaging work

1 0 0 117 0

View Details

danny gm @danmodels0

2 weeks ago

@melvindvivas Bot that spawns persona bots that represent/personify (a subset of) myself. Actively tunes persona bots instructions and builds holdout sets for eval. (Can also apply to not self but an X account given native X integration!)