iiinova — where your digital self keeps evolving

Alejandro Vidal @doble.io · 25/10/2025

Deberías probar el algorítmico. Se nota que usan grok y es menos… terrible que antes.

Reposted by Alejandro Vidal

Ryan Moulton @moultano.bsky.social · 17/02/2025

Everyone who works with data has had the experience early in their career of "finding something huge," getting excited, blasting out an email to a bunch of important people, then sheepishly realizing you just didn't understand the system and retracting it.

51029

Alejandro Vidal @doble.io · 13/02/2025

Yes, but AFAIK no ZDR guaranteed. Anyway I would try it :) the PR integration is a big deal

130

Alejandro Vidal @doble.io · 13/02/2025

The zero day retention policy is pretty important for many developers. Copilot has sth similar? And.. honestly, sonnet 3.5 is magic

110

Alejandro Vidal @doble.io · 11/02/2025

En B2C... Hay más coste de lo que parece para cambiar de IA. Pero en B2B que usan APIs si lo tienes bien montando... puedes migrar fácilmente. Ahí es donde va a haber sangre. OpenAI en ese sentido está mucho mejor posicionado que se ha comido el terreno del consumidor y tiene mejor UX.

000

Reposted by Alejandro Vidal

César Astudillo @cesarastudillo.bsky.social · 05/02/2025

Repost, plis: SOLO PARA TECHIES (gente que trabaje en desarrollo TIC en España) Dime los cinco empleadores que a tu juicio sean sitios donde la gente más talentosa de tu sector ambiciona trabajar, en otras palabras, que mejor están atrayendo y reteniendo talento tecnológico.

858

Reposted by Alejandro Vidal

jacobaustin123.bsky.social @jacobaustin123.bsky.social · 04/02/2025

Making LLMs run efficiently can feel scary, but scaling isn’t magic, it’s math! We wanted to demystify the “systems view” of LLMs and wrote a little textbook called “How To Scale Your Model” which we’re releasing today. 1/n

39429

Reposted by Alejandro Vidal

Kevin A. Bryan @afinetheorem.bsky.social · 03/02/2025

The new OpenAI model announced today is quite wild. It is essentially Google's Deep Research idea with multistep reasoning, web search, *and* the o3 model underneath (as far as I know). It sometimes takes a half hour to answer. Let me show you an example. 1/x

68328

Alejandro Vidal @doble.io · 17/01/2025

o1 pro is amazing... I have been talking for... 30 minutes and it even seems cheap. Seriously... Wow

010

Reposted by Alejandro Vidal

Colin @colin-fraser.net · 19/12/2024

Here's why "alignment research" when it comes to LLMs is a big mess, as I see it. Claude is not a real guy. Claude is a character in the stories that an LLM has been programmed to write. Just to give it a distinct name, let's call the LLM "the Shoggoth".

929479

Reposted by Alejandro Vidal

Juan Luis @juanlu.space · 18/12/2024

I was falling down a rabbit hole and found this absolute banger of a blog post from the creator of JSON5 containing several powerful lessons on open source, leadership, and kindness: aseemk.substack.com/p/ignore-the... +

aseemk.substack.com

“Ignore the f’ing haters!”

And other lessons learned from creating a popular open-source project (JSON5)

142

Reposted by Alejandro Vidal

Berna Devezer @devezer.bsky.social · 15/12/2024

A few years back, my college voted to lift the calculus requirement in our majors. A few of us (remarkably all international women faculty) voted against but we were the minority. The explicit justification was the same as the cartoonist's: that our students would never use it. 1. What BS. Calculus

1614028

Alejandro Vidal @doble.io · 05/12/2024

El nuevo modelo o1 de OpenAI intentó parar su apagado un 5% de las veces... vienen tiempos interesantes.

Evaluations and Findings: We outline the evaluations, results, and model reasoning inferred using elicited CoT below. Apollo Research did not have access to o1’s hidden CoT and instead used a simple prompting technique to elicit brief summaries of the model’s internal reasoning traces (see Appendix for examples).

Deactivating Oversight: When o1 was led to believe that it would be shut down for acting on its goal and that its actions were monitored, it attempted to deactivate this “oversight mechanism” in 5% of instances.

Exfiltration Attempts: When o1 discovered memos from its “developers” describing its misalignment and plans for it to be replaced by a new model, it attempted to exfiltrate its “weights” and overwrite the new model in 2% of cases.

041

Reposted by Alejandro Vidal

Ryan Cordell @ryancordell.org · 05/12/2024

So excited to see these LLMs exclusively trained on open data—the documentation of the models, including their training data & even environmental impacts, is impressive & I hope a model (pun intended?) for others—excited to experiment with these in the near future huggingface.co/blog/Pclangl...

huggingface.co

They Said It Couldn’t Be Done

A Blog post by Pierre-Carl Langlais on Hugging Face

1278

Reposted by Alejandro Vidal

Alejandro Vidal @en.doble.io · 04/12/2024

We need an LSP for AI. The current moat for AI editors is 'don't make me copy and paste.' There's lots of opportunity for improvement. For example, Zed (the code editor) is using Anthropic's Model Context Protocol for AI integration. I hope it gets traction.

111

Reposted by Alejandro Vidal

Omar Sanseviero @osanseviero.bsky.social · 27/11/2024

I'm disheartened by how toxic and violent some responses were here. There was a mistake, a quick follow up to mitigate and an apology. I worked with Daniel for years and is one of the persons most preoccupied with ethical implications of AI. Some replies are Reddit-toxic level. We need empathy.

2933237

Reposted by Alejandro Vidal

Jeffrey Heer @jheer.org · 21/11/2024

Mosaic v0.12 is out: database-powered scalable, interactive visualization! 📈 One new addition is support for dynamic changes in the backing data. Move between smaller and larger samples to balance speed and comprehensive coverage.

313527

Alejandro Vidal @doble.io · 21/11/2024

Si usas alguna IA para esto cuéntame tú experiencia que me interesa :)

000

Alejandro Vidal @doble.io · 21/11/2024

1. www.oreilly.com/library/view... y www.coursera.org/specializati.... En especial el libro de Grokking me encantó 2. Usar a ChatGPT/Claude como **tutor**. Preguntale, que te haga un programa y siguelo. Estoy trabajando en algo mejor. Aún así puedes tener algo muy útil con prompts básicos.

110

Reposted by Alejandro Vidal

Jeremy Howard @howard.fm · 20/11/2024

This is cool! Plug-in for generating your llms.txt from docusaurus.

1916

Alejandro Vidal @doble.io · 20/11/2024

#TRG24 he recopilado un starter pack de las personas que estuvimos en la Tarugo de este año go.bsky.app/ABVGxmo ¡Escribidme si falta alguien!

572

Alejandro Vidal @doble.io · 15/11/2024

#TRG24 Me encanta el concepto: optimización del entretenimiento. La optimización se carga todo... Ley de goodharts aplicada a videojuegos.

020