Ohhhh very nice.
We should figure out a good game to play with it. Like adding transport infrastructure or you need to achieve X time reduction in commuting times.
Nice!
28 karma · joined June 28, 2022
We should figure out a good game to play with it. Like adding transport infrastructure or you need to achieve X time reduction in commuting times.
Nice!
Check https://www.pleasedontdeploy.com/p/the-case-against-ai-works...
Tbh, with the swarm, we cover for that particular need. The caveat is that we don't try to be the SOT. At the end of the day, your company knowledge is distributed by design, so you're system needs to account for that. Instead of trying to have everything up to date, you need to remember where to look for the knowledge... even if it means burning a bit more tokens.
If so, why didn't you choose any of them?
I can tell you our experience, but I'm prob biased.
Also, boca o river?
Particularly interesting for you, our lasts posts on how we handle to have efficient dreaming focused on reducing costs.
tl;dr: there's a self-driven process generating more effective and efficient prompts, code, and potentially select systemone models to tackle the cost problems.
We have companies using the system that we understand have stayed 'flat' in token consumption but with increasing workloads (ie. more workflows running).
Hope it helps!
I wonder what would be the way to protect OSS like this from this kind of end?
It's a great win for the team, but many will be looking into other solutions for a while.
-> Clairic extracts the PR's claims -> Jev tags diff hunks, then judges what requires a review. -> Cheap(er) model to review the hunks.
For us, we've got PR reviews in under 30 seconds, and cost down to a few cents.
Next step, we are thinking on training some small open weight models specialized on our repo.
Thoughts?
You could guess, but would that change something?
If you introduce too many changes at the same time, how can you attribute what works and what not?
That's the other challenge we are seeing, and that gets worse if you have bigger teams.
Centralized intelligence helps, but enforcing the correct coordination... feels like we are missing a new type of role.
DeepSeek caveats: + You have an LLM-Agent-UMF architecture. Vanilla pi harness fails. + Coding tasks are lousy. Beyond coding, difficult open-ended reasoning scenarios are the ones at which it performs best for us.
It would have been much better if it was a skateboard, tbh.
Fun stuff!
Eventually providers that host models with tight restrictions will become popular just because of this. Everyone is collecting data, one way or another one.
What's funny for me? -> Companies are still signing multi-year contracts with big labs so they can send all their data there. -> If I was a big lab, I just need to learn behavior, not necessary even understand the messages, to take my software to the next level.
But who knows, I'm no genius.
So many people discussing things like UBI, however we selfishly create our own little projects all the time.
We need to center our shared efforts, send open source is a step on that.
Nowadays, every closed source solution should be seen as 'you are the product' type of deal.
This together with bombadil (web version pbt / Hegel / antithesis) for qa is a great advance.
We need more and more solutions like these for Agentic Coding.
I wonder if a micropayment would be detriment enough for most, even if it's just 1ct.
Above all, I'm also surprised on how those same organization are using Anthropic or OpenAI or other close source solutions for their agent harnesses instead of going for Open Source.
Malte just yesterday showed how powerful innovation with small teams can be achieved particularly in EU.
I hope they start looking for those alternatives too for their agentic systems, beyond using pi-mono.
I wonder how many here are considering that idea.
If you need determinism, building atomic/deterministic tools that ensure the thing happens.
I agree a framework is something that sounds outdated.
I also believe an orchestrator is needed. Something that abstracts you from a specific provider. Like hardware, drivers and operating systems.
Right now, my thoughts are on that line: Who will build that operating system? Who will have it in the cloud?
It needs to be robust to operate for large organizations, open source, and sit on top of any provider.
Right now we are seeing BSD vs GNU/Linux vs DOS kind of battles.
Above all, with everything that's happening in the software engineering world rn, I look at Chess as a place were we've seen it play out in the past decades. And Lichess is a big part of that.
I hope this deal helps two things: (1) Bring more people to Chess, (2) Actually, help Lichess find out a way to reward those working in it as much as they deserve.
Keep on the amazing work,
We also added pi-mono, and started using more and more other models for different tasks (Gemini, K2.5, GLM-5, you name it).
I think the problem is that most are building solutions that rely in one provider, instead of focusing self learning capabilities on improving the cost-quality-speed ratio.
For reference: https://github.com/desplega-ai/agent-swarm
ADK was (and is) exceptional, but nobody is actually making noise and pushing for it as they should. It feels like Microsoft .net back in the day.
Let's see how it goes. I'm rooting for y'all
I'll give sonnet another go.