Diogo Almeida

Sane + 🌶️ takes in an insane AI world... AI capabilities researcher: co-created RLHF/ChatGPT @ @openai now trying to right the wrong 🤭 (ceo @typesafeai)

[Diogo Almeida @CompleteSkeptic]

Happy JevDay!

But before announcements we have to talk philosophy…

The needs of early adopters / tinkerers / dreamers are quite different from what is needed for extremely stable infrastructure.

For the Jev release, we erred on the latter, but this release is for the former - with a gate so people can opt-in to the danger zone!

[Diogo Almeida @CompleteSkeptic]

@IgorIlyinsky we/I didn't make the site

[Diogo Almeida @CompleteSkeptic]

a crazy rethinking of spreadsheet ux: https://t.co/2mDjmmDwaS 🤯

use jev to semantically highlight cells

(we were trying to figure out sick excel demos for a while and this is way sicker)

[Diogo Almeida @CompleteSkeptic]

@gooby_esq @typesafeai agree. secret. 🤐

[Phil Renaud @phil_renaud]

Hi! I've built a Chrome extension that uses a decision model to turn every table on the web into a data visualization, so you can see patterns at a glance.

The decision model (@typesafeai's Jev) is perfect for solving a big hiccup with doing this: sometimes implied order of a column isn't "big numbers = good, small numbers = bad". Consider Earthquake deaths, or cost of a GPU, or Pitcher ERA. Jev is REALLY good and fast at figuring this stuff out.

[Erik Spock Gafni (hiring!) @EGafni]

Love the framing of "insert common sense here"

[Diogo Almeida @CompleteSkeptic]

WANT 🥵

(I think voice interfaces will be soooo huge)

[Kunal Abichandani @kunalabcdani]

What if we could manipulate UI at the speed of thought?

Built this voice coffee ordering system with @typesafeai Jev. The drink and any updates show up while I’m still mid-sentence, ~200 ms from word to screen.

Feels pretty magical!

[Diogo Almeida @CompleteSkeptic]

another jev powered launch! 🫶

[Hamza Tahir @htahir111]

Five years ago @adamvpro and I started @zenml_io because getting AI from a notebook into production was painful.

It still is.

Only now it's agents, not just ML models, that make it hard.

Building an agent takes an afternoon. Keeping it working is the hard part. A prompt tweak breaks something that worked yesterday. A cheaper model quietly stops answering. You only hear about it when customers complain.

Every team we talked to said the same thing: we don't need another dashboard, we need someone to do the work.

So we built that.

ZenML Pro now comes with an always-on AI engineering team. It reads your traces, finds what your evals miss, and proves the fix works before you ship it. This tech is backed by ZenML and Kitaru, our two OSS products, that stay open source and freely available.

To kick off this new chapter, we have decided to rebrand. ZenML is now ZenML Labs, and our ambition is greater than ever.

Connect your agent traces today, free at https://t.co/gN9rwIELjz.

We are very proud of the onboarding ;-)

We wrote about it here: https://t.co/BvCmE1fQB7

[Diogo Almeida @CompleteSkeptic]

yes! sounds about right

> the plumbing sequence of text based eval/field compiling: [copying fields, compiling JSON fields for API POSTS, writing metadata] That's >half our playbook, and Jev doesn't fit that typology. It stays with the LLM for now.

users have found success still using a string LLM for this, but using jev as a correctness classifier on top

> we're scoping what a Jev-first version looks like

this is 100% right! it often takes re-architecting to make the most of the new paradigm

[Diogo Almeida @CompleteSkeptic]

always good to continually learn from @sedielem

[martin_casado @martin_casado]

The legacy of Bostrom is going to haunt us forever.

This is all so boneheaded and avoidable.

[Diogo Almeida @CompleteSkeptic]

@swyx

[Diogo Almeida @CompleteSkeptic]

@swyx

[Diogo Almeida @CompleteSkeptic]

DOES THAT SAY SYSTEM 2 ON THE STREETS SYSTEM 1 IN THE SHEETS? @swyx are you implying what I think you're implying? 👀

[swyx @aiDotEngineer CODE @swyx]

sticker game inspired by @dexhorthy @CompleteSkeptic et al

[Diogo Almeida @CompleteSkeptic]

good thing none of the models are benchmaxxed, right? 😬

[Séb Krier @sebkrier]

Updated version!

[Gergely Orosz @GergelyOrosz]

How the hell does Jev cost $0.00 after using more than 1M tokens (on my new search site)

an LLM would have charged me $3-5 already. Jev pricing is incredible, pricing is per *billions* of tokens not millions

huge reason this spreads like wildfire. great work @CompleteSkeptic

[Diogo Almeida @CompleteSkeptic]

@tejasmanohar @typesafeai i reposted too, but just want to say thanks for the detailed report - will help us make jev better 🙏

[Tejas Manohar @tejasmanohar]

our eng team got #1 on HN today!

they tested if fast + cheap decision models like Jev @typesafeai could grade complex data-matching tasks as well as frontier models.

Jev with Claude Opus got close to Opus-level accuracy for a fraction of the cost.

super cool. post below:

[David Luzar (dwelle) @dluzar]

@AndrewCurran_ My girlfriend goes to a different compute campus.

[Diogo Almeida @CompleteSkeptic]

@sedielem 😠🔪

[Diogo Almeida @CompleteSkeptic]

how @sedielem triggered me at dinner

[Aidan Clark @_aidan_clark_]

@immortaldip Everyone who actually works at frontier labs understands that data doesn’t matter.