Manuel @ManuelSH

Accelerating the adoption of AI at scale. manuelsh.github.io Joined February 2009

Tweets

768
Followers

153
Following

1K
Likes

3K

Manuel @ManuelSH

2 months ago

@levie That’s the Jevon’s paradox: “is said to occur when technological improvements that increase the efficiency of a resource's use lead to a rise, rather than a fall, in total consumption of that resource”

0 0 0 46 0

View Details

Manuel @ManuelSH

3 months ago

AI maturity comes in levels: Level 1: Write better prompts. Iterate in the conversation. Level 2: Context engineering. Ensure the model has the right info at the right time. Level 3: Harness engineering. LLM has the right sources, tools, feedback loops, safeguards, etc

0 0 0 32 0

View Details

Manuel @ManuelSH

3 months ago

@theprototype003 @DotCSV Sí, yo lo veo así. El frontend lo generará la AI. De hecho, aquí hay una prueba de concepto... x.com/zan2434/status…

Zain Shah @zan2434

3 months ago

Imagine every pixel on your screen, streamed live directly from a model. No HTML, no layout engine, no code. Just exactly what you want to see. @eddiejiao_obj, @drewocarr and I built a prototype to see how this could actually work, and set out to make it real. We're calling it

1K 4K 29K 6.0M 25K

0 0 0 26 0

View Details

Manuel @ManuelSH

3 months ago

Such a good reflection

Ethan Mollick @emollick

3 months ago

Every AI discussion ultimately rests on two questions: how good can AI get? And how fast? They are predictions about the s-curve shape. Everything else (job impact, potential risks, etc.) is downstream of those questions. I think it would be useful to focus on them more often.

64 67 462 29K 171

0 0 0 46 0

View Details

Manuel @ManuelSH

3 months ago

Here it is: github.com/manuelsh/cv-bu…

0 0 0 22 0

View Details

Manuel @ManuelSH

3 months ago

Wanted to update my CV... but ended up building an app that orchestrates multiple LLMs to generate a custom CV for each situation. And of course I built it with an agnet: @OpenAI Codex.

1 0 0 44 0

View Details

Manuel @ManuelSH

3 months ago

@emollick The key that will change everything is continuous learning. x.com/ManuelSH/statu…

Manuel @ManuelSH

3 months ago

In the future, we’ll remember context windows as a strange, almost funny constraint we once had to live with.

0 0 0 222 0

0 0 0 173 0

View Details

Manuel @ManuelSH

3 months ago

In the future, we’ll remember context windows as a strange, almost funny constraint we once had to live with.

0 0 0 222 0

View Details

Manuel @ManuelSH

3 months ago

OpenClaw seems one, but looks like a proof of concept. ChatGPT and Claude has 3 or 4 of all the points above, but not all. Same with Claude Code and Codex. But still not a finished product.

0 0 0 59 0

View Details

Manuel @ManuelSH

3 months ago

The most obvious AI product still doesn’t exist: a personal agent (1) connected to email, messaging, the internet, my files, my tools... (2) that keeps context about me (e.g. see LLM Wiki), (3) available from any interface, (4) proactive, and (4) secure. We have glimpses of it.

1 0 0 59 0

View Details

Manuel @ManuelSH

3 months ago

@philo_societycj @411emilio Same thing was used in ancient Rome, where daytime and nightime was divided in 12 “hours”

0 0 0 24 0

View Details

Manuel @ManuelSH

3 months ago

@OpenAI @nikey_es lo que hablamos justo hoy…

0 0 1 14 0

View Details

Manuel @ManuelSH

3 months ago

@javilop ¡Me alegra ver que “autorreason” parece funcionar!

𒐪 @SHL0MS

4 months ago

@karpathy i've been working on a method called autoreason that is effectively autoresearch extended to subjective domains. autoresearch works because val_bpb gives you an objective fitness function. autoreason constructs a subjective one through independent blind evaluation, the same way

24 12 372 202K 440

0 0 0 142 0

View Details

Manuel @ManuelSH

3 months ago

@emollick Indeed. Agents can easily compound LLM hallucinations.

0 0 0 72 0

View Details

Manuel @ManuelSH

3 months ago

@karpathy @existeundelta Very nice! In some ways is similar to what I am building, with more human curation: theoria-dataset.github.io/theoria-datase…

0 0 0 14 0

View Details

Manuel @ManuelSH

4 months ago

Almost everyone wanted Cursor. Very few wanted Claude Code. Our evaluation pointed the other way: Claude Code performed best. Good reminder that in AI tooling, popularity is not the way to choose. adevinta.com/techblog/a-pra…

0 0 0 88 0

View Details

Manuel @ManuelSH

5 months ago

@DotCSV 14 feb 2019. Me dije: “ya está resuelto”.

0 0 0 424 0

View Details

Manuel @ManuelSH

7 months ago

A great resource…

elie @eliebakouch

8 months ago

Training LLMs end to end is hard. Very excited to share our new blog (book?) that cover the full pipeline: pre-training, post-training and infra. 200+ pages of what worked, what didn’t, and how to make it run reliably huggingface.co/spaces/Hugging…