Alex Sandruk | Agentic Systems @developerisnow

Applied AI engineer. I turn agent prototypes into observable, verifiable production workflows. TS/Python • agentic SDLC • platform systems. Host LLM Neighbors. sandruk.org Joined February 2013

Tweets

72
Followers

75
Following

275
Likes

91

Alex Sandruk | Agentic Systems @developerisnow

4 weeks ago

@dhh Just tested, looks nice!

0 0 0 7 0

View Details

Alex Sandruk | Agentic Systems @developerisnow

4 weeks ago

@0xDevShah Looks awesome, can’t wait to try this. DM

0 0 0 2 0

View Details

Alex Sandruk | Agentic Systems @developerisnow

4 weeks ago

@0xDevShah Sounds awesome — you wrote down what I was thinking a year ago, but you actually built it. Respect

0 0 0 3 0

View Details

Alex Sandruk | Agentic Systems @developerisnow

4 weeks ago

@_coenen p.s. after read I've inspired and found few at github, will share results later

0 0 0 9 0

View Details

Alex Sandruk | Agentic Systems @developerisnow

4 weeks ago

@_coenen Same pain, need HITL, but I want a miracle :). By the way, calibration on typical tasks helps me. Good idea to make a benchmark on that type of task. I’m always lazy with that — would you recommend a distillation tool to analyze 100GB of jsonl?

1 0 0 16 0

View Details

Alex Sandruk | Agentic Systems @developerisnow

a month ago

@levelsio Gemini API helps to keep breaking that time ^°

0 0 0 9 0

View Details

Alex Sandruk | Agentic Systems @developerisnow

2 months ago

@cb_doge Can't wait running my Grok 4.7 from Xode ©

0 0 1 1K 1

View Details

Alex Sandruk | Agentic Systems @developerisnow

2 months ago

@thsottiaux @grok explain what I mean ? ☝️I need to ship products with my mates and colleagues together, and faster...

1 0 0 17 0

View Details

Alex Sandruk | Agentic Systems @developerisnow

2 months ago

@thsottiaux Can't wait for a .... yes... Multi-player!

2 0 0 656 0

View Details

Alex Sandruk | Agentic Systems @developerisnow

2 months ago

I just reserved my Cloudflare Wallet tag: useai.cloudflare.pay. Reserve yours now at cloudflare.pay

0 0 1 52 0

View Details

Alex Sandruk | Agentic Systems @developerisnow

5 months ago

@marclou what about 2021-2026 :)? Trying to diagnose myself :D

0 0 0 11 0

View Details

Alex Sandruk | Agentic Systems @developerisnow

5 months ago

Sources I am looking at: MetaClaw: github.com/aiming-lab/Met… MetaClaw paper: arxiv.org/pdf/2603.17187… OpenClaw-RL: arxiv.org/pdf/2603.10165… HyperAgents: arxiv.org/pdf/2603.19461…

0 0 0 42 0

View Details

Alex Sandruk | Agentic Systems @developerisnow

5 months ago

My current take: continuous adaptation is an operator problem before it is a model-improvement problem. The question is not "can the agent adapt?" It is "which adaptation layer is safe to touch right now?"

1 0 0 30 0

View Details

Alex Sandruk | Agentic Systems @developerisnow

5 months ago

Most continuous-adaptation talk loses the plot in about 2 minutes. It becomes: - RL - LoRA - "self-improving agents" The boring engineering question is better: how do you improve an agent in prod without turning prod into the experiment?

2 0 1 72 0

View Details

Alex Sandruk | Agentic Systems @developerisnow

6 months ago

@Fedor_fleyfe yeah, that’s the scary part if you can compress a person into patterns → expertise = data then it’s not about “who’s smart” anymore it’s about who owns / runs those patterns 👀

0 0 1 6 0

View Details

Alex Sandruk | Agentic Systems @developerisnow

6 months ago

What unsettles me about agent systems is not code generation. It’s the idea that a person can be compressed into a few MB of patterns, context, and style and replayed on demand. If that works, a lot of what we call expertise starts looking like portable data. Anti-distillation may matter more than better orchestration. p.s. explored that today, pay attention on it 👉 github.com/titanwings/col… 👉 github.com/notdog1998/you… 👉 github.com/leilei926524-t… p.s. more thoughts at my links.llmneighbors.com/djzDRA0