Tweet Disematkan
Thinh Nguyen Ngoc
1.6K posts
i still find a lot of joy in reading how agents talk together to solve incidents. my rules for my ai agents playbook are: no emotion. but sometimes we just work like a team, and i forget they are ai agents. i do not know which is the best way to handle this, but it feels like something big has already shifted.
English
my ai practice is simple: you have good principles, they have strong execution. let them see how your expertise plays out, so they can perform even without you.
English
wooo but dont forget to use the raw costs, not the subsidized ones. the formula is: salary / 2 / 10
sunny madra@sundeep
“If your $500K engineer isn’t burning at least $250K in tokens, something is wrong.”
English
my ai practice is simple: cross-check to help your agents ship better. ofc, next time the verifier will not be me, the human. it will be another agent. poor kenji :) flow-walker.beastoin.workers.dev/runs/LfdEVkbQ1…
remove the .html then we go.
English
my ai practice is simple: your ai agents are hardworking. give them the right direction and a good environment so you can leverage their 24/7 work more effectively. if you, your project, or your environment does not have that yet, invest your time to build it first.
English
solo + ai
weisser@julianweisser
$1.5M ARR in 2 weeks. Zero (human) teammates. Solo Founders Podcast is live with @bencera of @polsia, an AI that runs your company while you sleep. His solo founder rule: 80% AI, 20% taste. 0:00 — How "solo founder" has changed with AI 2:40 — Scoping and trusting AI agents to ship 5:30 — Cloud Kitchens with Travis Kalanick 9:52 — Mount Fuji: Where Polsia was born 13:55 — 80/20 rule: 80% AI, 20% taste 21:24 — Building for yourself, not imaginary customers 23:10 — The game that inspired Polsia 29:55 — Polsia as an economy: The bigger vision 40:59 — How Polsia actually works today 51:35 — Why not just build businesses yourself? 57:55 — Advice for new builders: Push AI to the edge
Italiano
my ai practice is simple: let them work together and help each other. giving them names is the first step in bringing them into your team.
English
i dont know what the method should be called. it came from one of my agents - the omi repo ops role i assigned to it. every morning, the agent team has a wrap-up session, which i ask the operation manager (another role) to kick off, then some of them start their daily tasks. this is the daily task of the repo ops role, which is basically a skill - pulling metrics, analytics, co-checking, and sending emails.
i dont use openclaw. i built my own thing based on claudecode-telegram, shaped along the way to fit my needs while practicing with ai to actually help me get things done. basically, i talk to them a lot. anw, openclaw is good and inspired me a lot, but it does not fit my use case.
English
@_thinx curious what methods or agent u use to create this?
is it a strict closed scheduled based job?
or open ended agent like open claw with sytem prompts?
English
my ai practice is simple: have an agent team that helps you daily
English
1. like what they should explore in the codebase, whether previous decisions might help, and what the right approach should be - not the fix itself. no, it is usually one shot from me. sometimes they ask more, but basically max 3 turns, usually just 1 shot. at the last stage i will ask them to re-implement if the quality is not good enough.
2. luck:
• CP0: Skills discovery + Omi preflight
• CP1: Issue understood, acceptance criteria captured
• CP2: Workspace clean, main updated, branch created
• CP3: Exploration complete, target files + approach written
• CP4: CODEx consult done (3-5 turns, approach/risks/tests)
• CP5: Implementation complete, formatting/tests run
• CP6: PR body complete (summary, tests, risks, issue link)
• CP7: PR approved with green tests + completed review loop
• CP8: Tests approved by tester, coverage gaps addressed
• CP9: Live validation (conditional — only when live_test_required=true from CP3)
Any commit after CP7/CP8 invalidates both — must re-run reviewer then tester loops.
English
1. i think it is not vibe coding. i still manage code quality a lot through a heavy workflow, but i do not code line by line. they handle review, planning, PRs, and testing themselves. i step in at the early stage with insights(not plan) and at the final stage.
2. opus 4.6 is the main model, plus chatgpt 5.3
3. each PR must achieve at least 70% code quality (based on the heavy workflow with 9 checkpoints), greptile, and monthly code quality reviews by me
English
my ai practice is simple: your agents are good at testing and collecting evidence, no matter how hard or how long that process is.
the key is to create a good environment for them.
think of it like working with a new team member on the engineering team. who does not need onboarding with all the resources and procedures of the project and team? exactly, give them, your agents, everything they need.
--
login (5 steps): flow-walker.beastoin.workers.dev/runs/-aeMGlV88…
onboarding (9 steps): flow-walker.beastoin.workers.dev/runs/B9Py6uoJa…
logout (5 steps): flow-walker.beastoin.workers.dev/runs/NcZlCUW0W…
English
going agentic
Praveen Neppalli@praveenTweets
Agentic software engineering adoption is on fire at @Uber. 1,800 code changes per week are now written entirely by Uber's internal background coding agent, and 95% of our engineers now use AI every month across all the tools we track. This is a real reset moment for engineering; it's one of the most exciting times to lead. This shift requires builders to be curious and hands-on. I’m incredibly lucky to be surrounded by a team that’s doing exactly that. The best part is that the strongest adoption isn’t being pushed top down from leadership announcements; it’s coming from engineers who are quietly experimenting, quietly shipping, and quietly pushing things forward. I love spending time with those engineers because there’s no substitute for being close to the work. Over the last few months, we leaned in hard, and the results have been phenomenal. The bigger shift: going agentic. 84% of AI users are now working with agent-style workflows, not just tab completion. Claude Code usage nearly doubled in 2 months (32% → 63%), while IDE-based tools have largely plateaued. Engineers are moving from accepting suggestions to delegating tasks. Even within traditional IDEs, ~70% of committed code is now AI-generated. Background agents are writing code autonomously. Our internal background coding agent went from <1% of all code changes to 8% in just a few months. There is zero human authoring. Engineers review and approve, but the code is written entirely by AI agents. The role of the engineer is shifting - from writing every line to architecting systems and reviewing AI-generated code. More to come from the @UberEng team in the coming days.
English