Register and share your invite link to earn from video plays and referrals.

Paul Bakaus
@pbakaus
Renaissance geek (creativity × culture × tech). Product engineer before it was cool. Made @impeccable_ai, @radiant_shaders, @jqueryui | x-@google, @zynga
682 Following    21.1K Followers
beware the sloptopus 👀 (this is a real screenshot)
most popular prompt today: “hey what could we use Jev for in our repo?”
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x cheaper (w/ output tokens free) • Frontier composable intelligence optimized for decisions AFAICT the shortest path to AI-based economic revolution
Show more
some interesting facts about Impeccable for PRs: - much more advanced version of "/impeccable critique" and "/impeccable audit" - slots right into your ai bugbot dev workflow (if you're already using Greptile, CodeRabbit etc, your agent will simply pick up on it and your UX auto-improves) - much more powerful than some skill that estimates what your code would look like when rendered. spins up real browser environments, logs in, looks at before/after... - AI repository analysis makes sure it understands your design system (and changes to it over time), and your target audience, and constructs simulated personas from it who test every new feature (yes, with actual voiceovers and videos!) we've worked on this for a couple of months and have been using it on our own internal repos for many weeks. it's @renaissgeek's first commercial product. very excited to have you try it.
Show more
engineering is moving 10x faster. design review hasn't caught up. introducing Impeccable ..for PRs! a design review agent that simulates your users, uses your app, finds friction & slop. Impeccable’s design judgment in every PR. reply if you want in on the closed beta 👀
Show more
@pbakaus keeps delivering and expanding @impeccable_ai's reach (and value). This time to PRs.
engineering is moving 10x faster. design review hasn't caught up. introducing Impeccable ..for PRs! a design review agent that simulates your users, uses your app, finds friction & slop. Impeccable’s design judgment in every PR. reply if you want in on the closed beta 👀
Show more
0
165
336
13
Forward to community
Product people of SF! On Sep 22 we're hosting an event with the incredible @pbakaus (Founder of @impeccable_ai) and @moeamaya (Founder of @operatecrm). If you care about creativity and craft and want to hang out with some of the best design talent in the Bay Area, don’t miss this one!
Show more
If you're interested in the origin story of Impeccable and the philosophy of it, here's my talk on the subject from @aiDotEngineer world fair a couple months back (finally online!):
Show more
The old Impeccable site was flashy. For a design toolkit, I wanted something calmer, cleaner, and more elegant. So I redesigned it, rewrote the docs, and added more ways to explore, learn, and experiment. Thanks @Abdul_wahab_one for helping me bring it over the finish line. Here’s a quick tour.
Show more
judging from the popularity of this tweet, it's great that at least a few humans seem to care about the faith of humanity! on the other hand, the discourse (and the other follow-ups) is braindead even for X' usual low bar, ranging from "it's very financially irresponsible to warn folks about skynet taking over right before skynet's IPO" to "you're dumb it's just next token prediction". i wish everyone would go through a quick thought experiment like "hey if AI has access to all our infrastructure, and suddenly isn't aligned to humans but maybe nature, the planet, or something like that, would that be bad?", which, i'd hope, would be lead to a pretty sobering 'oh.'
Show more
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
Show more
GPT 6 Astra's software development's performance (super spiky/random) is probably the best case for graph engineering I've seen yet. Run the same challenging task 5x, and you get: - 3x: a stick figure - 2x: god-tier Michelangelo So do that, then let a 5.6 judge pick the best.
Show more
looking forward to this one!
Another edition of @Dessn_ai's Creative Machines - this time in SF! Kicking off with a fireside with two incredible, high-craft founders, @moeamaya and @pbakaus 🤩 Followed by fun trivia and a demo from our lovely users at @Color to finally see @k_grajeda's beautiful private-beta work in action. Spots are limited. RSVP link in the comments.
Show more
oh hi @NousResearch! if you missed it in the long 4.1 changelog: Impeccable is now officially supported in Hermes Agent (and thx to @Teknium for helping update our docs via a PR!)
this is very cool but the voiceover is incredibly…um…optimistic. no, this will not replace the need for stable user interfaces. cute demos like these dramatically misunderstand, or worse, belittle UI design. and this is commentary from somebody who loves the vision of generative ui (I own the domain
Show more
Today, we're sharing new research on Solaris, our first Interface World Model. Solaris is a new kind of operating system that generates interactive interfaces frame by frame, in real time, with no code. We find that Solaris outperforms frontier LLMs when generating new interfaces, across structural similarity and information retention. Read more and request early access at the link below.
Show more
those who try to shame you for attempting inbox zero live a pitiful existence there’s nothing more critical for mental load than closing open loops, and every read-but-unanswered email is an open loop that will live somewhere in your brain rent free. inbox zero ftw
Show more
Welcoming Paul Bakaus @pbakaus, founder of Renaissance Geek @impeccable_ai, and a product engineer before it was cool. Speaking at Abstract Conference. Sept 2nd
Show more
this is one of my highly unpopular oss projects that i’m personally using even more often than @impeccable_ai 😅
what’s interesting about this is that it doesn’t really work often, especially not for frontend development. there is a hard ceiling for what a model like gpt 5.6 can accomplish, and when it realizes that it cannot move the needle on a harsh judging agent, it will simply give up.
Show more
Scaling self-verification with DeepSeek V4 Flash beats Claude Fable 5 on Terminal-Bench 2.1, while being 11x cheaper 💰 As open-source models become more capable, they can now generate large numbers of high-quality candidate solutions and verify their own outputs at very low cost. For example, we find that sampling just 5 solutions with DeepSeek V4 Flash and ranking them using the same model with LLM-as-a-Verifier can lead to a significant boost in accuracy (79% → 88%), outperforming closed frontier models on Terminal-Bench. Try it out today: More on verification scaling in my previous post.
Show more
how do design tools like claude design and other newer canvas based tools, the new /design command compare with @impeccable_ai? this is a common question, so i figured i’ll provide you with an answer: i see claude’s design tooling primarily used as a prototyping tool for early exploration, or for designing one-off documents. Impeccable, on the other hand, was born because I needed a daily driver to help me expand and improve the design in a large production web app. these are very different jobs to be done. many impeccable users prototype something in a canvas or claude artifact, and then use impeccable to finalize and iterate on their designs.
Show more
i’m unreasonably happy about this
You probably hate tab switching. Understandable. You can now open any @googledrive Doc, Sheet, or Slide right inside ChatGPT and work side by side without switching tabs. Rolling out on web to Plus, Pro, Business, and Enterprise users in ChatGPT and ChatGPT Work.
Show more