max

@kylemathews Ohh didn't know that electric agents will be part of the OSS??

@maxjendrall github.com/electric-sql/e… for Electric Agents

Self-modification has been achieved internally

@LizzieTao @gmail Gmail or also Workspaces? If Workspaces, the AI bubble is a good place to hook into

in a terrifying turn of events I now run social for @gmail! (great power, great responsibility, blah blah blah) let us know if there's anything specific you'd like to see from that brand!

@KosmynkaOS The link + the mention are both not existent

@badlogicgames @mitsuhiko @badlogicgames and @mitsuhiko thank you so much for joining!! @slobkebap and me where very happy you were able to make it.

I am glad to report we had @badlogicgames (while driving through Austria) and @mitsuhiko (while doing a BBQ) join us virtually for our first pi.dev Berlin Event. You can watch the full Q&A, as well as the demos here: youtu.be/BQ_Es8k650I?si… (starting at 23min)

@alex_whedon Already done, would love to try it!!

@maxjendrall Sign up at SubQ.ai, we'll get you in.

Introducing SubQ - a major breakthrough in LLM intelligence. It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA), And the first frontier model with a 12 million token context window which is: - 52x faster than FlashAttention at 1MM tokens - Less than 5% the cost of Opus Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention). Only a small fraction actually matter. @subquadratic finds and focuses only on the ones that do. That's nearly 1,000x less compute and a new way for LLMs to scale.

@lumendriada oh nice, when?

herdr crossed 400 stars 🎉 i’ve been dogfooding it over ssh to my vps/desktop, managing agents while away, and realized native tui alone isn’t enough. so here’s a sneak peek of what’s coming: web-like responsive ui. no separate app. still terminal-native.

@aliouftw @AmpCode Check control.ts from @mitsuhiko

was about to make a pi.dev extension to communicate between sessions but i actually only needed @AmpCode 's read thread tool:

@thdxr Oh I'd be curious why, please help us understand

i'd rather lose $100,000 a month with a team of 10 i feel sorry for you if you don't understand this

@thruflo you should make it a habit to also put a write up link in

Agents are not compute. Agents live in the data layer. Multi-agent systems are a sync problem: youtu.be/OiTqcScOFgE

Hey @thsottiaux @sama @benedictk__, I accidentally signed up to the 5.5 luma event with an email that does not have an openai account associated to it. Is there any way to get the 10x codex boost applied to my actual Pro account instead? 🥹 AI support says no Please

@andreasklinger @mayukh_panja Applied, please let meeeee innnnnnn

@mayukh_panja Tech twitter sounds super boring We do a 200 people meetup about robotics- come by luma.com/0devxzvp

We need a twitter tech meetup in Berlin. - small event, restricted to 50 people - very low entry fee: 10 euros ( free events bring in unserious attendees, people looking for free pizza types) - serious builders only - 5 talks. 10 min each + 5 min for Q&A Most events I see are too broad and - almost anyone can come - or extremely high entry fee (like 300+ Euros)

@mayukh_panja errr, no need to pay for it... @slobkebap and me are organising some nice events

@PsudoMike @spacetime_db Yeah,... I have the case where I am building a lot of software for customers where I just can't put them on a tech where I am not sure if it'll actually continue to support the web case.

@oorusr @spacetime_db Extensions plus the ops muscle around postgres is the moat. You can spin up a clever DB in a weekend, but the page at 2am when it locks is brutal. PG has 30 years of writeups and people who already had your outage.

@spacetime_db unfortunately fell of my radar despite the initial hype around the crazy demos. I did use it initially and was very very impressed but migrated to postgres on all projects where I tried it as I needed 1. the extensions and 2. stability and trust. Kind of unfortunate

@filearts Repo?

My first pi extension: /feedback. For the novels that models tend to write. It writes the last msg to a .md in your session and opens it in $EDITOR. If you save it with changes, it injects a [Feedback] token. The token is replaced with a small prompt + the diff. Works great!

@mitsuhiko @dmosher @badlogicgames @nonsens3 @mikker Oh damn, this also just happened to me. Would be nice to have some sort of context on the error message when it happens that it switching to SSE is the way to continue.

@dmosher @badlogicgames @nonsens3 @mikker I think the frame limit on web sockets is 4MB or roundabout that. I also think that node does not support permessage-deflate, so we probably don't have compression going.

pi.dev's been supporting OpenAI's WebSockets mode since March. /settings > transport. the difference between SSE and WS for e.g. Spark is not that stark in terms of tokens/sec. the real speed up actually comes from caching context OpenAI side, and only sending deltas as the context grows. pi currently does not implement the latter. hmmmmmm

@morgane_paloma @F1 @i_mohanram @braintrust @slobkebap lmao

being resourceful is underrated. i mentioned the traffic getting to @F1 was bad and the next day @i_mohanram had a police escort waiting for the @braintrust group. i didn’t even think that was an option. 💙