tarasyarema

6 ms·

1.
by tarasyarema 2mo ago
Hello all OG here. For anyone coming back to check the comments, thanks a lot for all of them and the feedback! I did a re-write of the post to remove the sloppy parts and making more clear some of the estimates and assumptions we made in t
2.
by tarasyarema 2mo ago
We run it ourselves since nov last year (~20k tasks since then, note we are 2 people and bootstrapped). Then we know a few companies using it w +1k tasks x day (~20/30 people actively using it daily). Then we have some anonymous teleme
3.
by tarasyarema 2mo ago
yes, typed "sdk"s are key, and indeed they are so good at js we did some iterations specially on the typing and error messages, as you mentioned, that help a lot in the reliability
4.
by tarasyarema 2mo ago
Love this guy, you should read the jokes book he has, pure gold
5.
by tarasyarema 2mo ago
exactly, I love how the cloudflare article puts it in fact
6.
by tarasyarema 2mo ago
I did not expect it either tbh :D But I believe it's not that common as it seems, only one actually trying to push this is executor.sh for example
7.
by tarasyarema 2mo ago
thanks for the feedback! it's really hard to measure some times. for example, once of the things we have seen is that the specific `workflow-triage` schedule we have have been running much faster and cheaper since we moved some parts o
8.
by tarasyarema 2mo ago
still most harnesses and AI systems do not use it, adopt it, or reference it enough...
9.
by tarasyarema 2mo ago
yes, that is true. in fact we are thinking a lot about this, not on a toll based, but rather on a role based. giving access to specific tools to specific agents, how that works and evolves, and how that links w the humans using it
10.
by tarasyarema 2mo ago
thanks!
11.
by tarasyarema 2mo ago
will do next time, thanks!
12.
by tarasyarema 2mo ago
yes, 100% but it's not that simple. the whole point of this is that if you have an AI system that is in charge of this, what you need is a way for it to build those deterministic software based parts, while maintaining the AI routing a
13.
by tarasyarema 2mo ago
This is based on a multi-day convo I had with our swarm. I might have not wrote it myself, but it's based on human interaction and thought process. Any recommendations on how I could make it better next time, so you do not feel like wa
14.
by tarasyarema 2mo ago
Yes, that's the right direction codex is taking 100% Now our idea is to offer this to be harness agnostic (we support codex, claude code, pi, opencode, even devin). And you (as the one deploying AI systems) should be in control of the
15.
by tarasyarema 2mo ago
I get it, but you got it right. It was assisted, not just AI slop. Also nice that it's not 100% :D
16.
by tarasyarema 2mo ago
we have a way to see created scripts in a dashboard, and in fact they are not ad-hoc only, i.e. it can create "saved scripts" that it can re-use, iterate and so on. Also not exec happens in a sandboxed env, which means it might no
17.
by tarasyarema 2mo ago
Hi, the blog post was written assisted with our own swarm, based on the experiment. Indeed we had the concept of workflows from the start (like n8n), but the point of the scripts is type safe executions that the agents can write, with acces
18.
Code mode yields a 99.2% cost reduction in our systems(agent-swarm.dev)
62 points by tarasyarema 2mo ago|69 comments
19.
by tarasyarema 2mo ago
Hi there! We've been building https://github.com/desplega-ai/agent-swarm in the open for a while, and one of the things we wanted to do since the start is find how powerful it can be to push for a code mode like e
20.
Show HN: The open-source alternative to Claude Tag(agent-swarm.dev)
3 points by tarasyarema 3mo ago|0 comments
21.
Slop Machines (2026)(tarasyarema.com)
2 points by tarasyarema 3mo ago|1 comments
22.
by tarasyarema 3mo ago
On why Polsia is the biggest slop machine we've seen to date.
23.
Claude -p drop-in replacement after Jun 15th(github.com)
2 points by tarasyarema 4mo ago|1 comments
24.
by tarasyarema 4mo ago
If you want to still use your anthropic sub after 15th of June, you could try this out. Still trying to figure out if there will be ways to "bot check" from Anthropic side for this workarounds.
25.
We are building an open-source agentic company OS(agent-swarm.dev)
1 points by tarasyarema 5mo ago|1 comments
26.
by tarasyarema 5mo ago
Hi there! For a while we've seen companies focused solely on the coding agent part, and specially building harnesses. We thought it would be interesting to go in another direction: building on top of them. Building on top of them allow
27.
Chat-py: A Python port of Vercel's chat SDK(github.com)
1 points by tarasyarema 5mo ago|1 comments
28.
by tarasyarema 5mo ago
Hey there HN! I loved the chat sdk for node, but I wanted to use it in Python, so I let my agent-swarm run in the night to port it (it only costed $500 in claude usage xD). Performing some E2E tests now, and will be releasing to pypi as `ch
29.
by tarasyarema 6mo ago
Hi there, welcome!
30.
Show HN: Agent-fs – The missing filesystem for agents (and humans)(github.com)
2 points by tarasyarema 6mo ago|0 comments

More ›