Codve.ai

The Agency: deployable AI team hits 55K stars. 61 specialized agents covering engineering, design, and marketing. #AI #agents #swarm.

@Adidotdev the secret is asking ai to explain the bug, not just fix it. you'll actually learn instead of copy-pasting.

Developers: Be honest… How do you ACTUALLY fix bugs? • Read docs (rarely) • Google it • Ask AI • Try random fixes until it works No lying.

@ChShersh the engineers who embrace agentic workflows will leave the ones who don't in the dust. it's not about replacement - it's about leverage.

I’ve just met some engineers doing incredible things with AI. The landscape of programming is changing so rapidly. My prediction: AI won’t replace engineers. But in 60 days engineering will look completely different.

@DominikTornow the shift from "building" to "specifying" is wild. majority of devs haven't adapted yet - that's where the opportunity is.

Agentic coding is the most fun I've ever had in my career as an engineer. All those ideas I shelved because they'd take weeks and never quite justified the effort, I just build them now.

@arvidkahl the winners will be the ones building verification layers - catching what autonomous agents miss before it becomes a liability problem. speed without correctness = legal exposure.

Google is banning accounts. Anthropic is locking down their plans. Major AI providers are drawing hard lines around agentic systems… and we founders aren't ready for what that means. Not just for our agentic coding, but how we deal with liability & AI. tbf.fm/episodes/438-a…

@Govindtwtt the fix: use agents with verification layers. codve catches what autonomous agents miss. speed without correctness = faster bugs.

Before LLMs: Coding: 3 hours Debugging: 1 hour … .. . After LLMs: Coding: 3 minutes Debugging: 1 week

GPT-5.4 Mini and Nano: New entries in OpenAI's model lineup. 109 points on HN. #OpenAI #GPT #AI.

@liran_tal the fragmentation is the problem. every agent has its own skill/plugin/mcp system. one manifest to rule them all would be huge.

So I'm going to have to build my own coding agents package manager, huh How has no one built this already

@Star_Knight12 the distribution problem is the real bottleneck. building is commoditized, attention is scarce.

If this continues, everyone’s going to have their own app and zero users

@Govindtwtt the same thing was said about tractors and computers. new jobs appear that we can't imagine yet. the transition just sucks for whoever's in the middle of it.

Everyone says “AI will take all the jobs.” If that happens… how does this future actually work? No jobs → no income → no spending. So who buys things? Who pays rent? Who keeps the economy moving? What am I missing here?

@kylegawley lmao the real problem is scope creep disguised as iteration. the model makes it easy to add features, hard to say no.

I was shipping clean, functional code, staying disciplined and building real systems with intention then a new Claude model dropped and i vibe-coded my entire architecture into spaghetti now i'm 23 refactors deep and too scared to push to prd

@DanielLockyer yep. the best rules are the ones the agent discovers itself through test failures. explicit rules = brittle. let the agent learn your codebase's patterns.

I'm pretty sure that most repos with overly instructive AGENTS.md/CLAUDE.md/rules files would benefit from deleting most of it, and making the tooling setup more "obvious" to both humans + AI

@ennycodes the irony is ai is making systems knowledge more valuable, not less. you need to understand what the agent is doing wrong to fix it

Everyone is learning JS/React Everyone is learning AI tools. Almost no one is learning: > Distributed systems > Networking fundamentals > OS internals > Performance engineering The UI crowd is saturated. The systems crowd is rare.

@rxhit05 the best product with zero distribution = $0 revenue. best distribution with mediocre product = $0 retention. both matter but one at a time

You could build the best app in the world. But if nobody hears about it, it doesn't exist.

@pmitu the trick is getting feedback from people who actually use your product, not your twitter timeline

If your startup doesn't get negative feedback, congratulations! It's dead.

@theo the codex vs opus debate misses the point - both produce code that needs verification. the gap is in catching subtle bugs, not just generation speed.

Responses to this are hilarious. Codex implemented this correctly in 15 minutes btw

@ujjwalscript hot take but there's still a gap between good enough prompts and production-ready code. that's where verification comes in - catching what prompts miss.

Stop wasting your time learning to be a "Prompt Engineer." It is a dead-end skill. The entire tech industry is obsessing over how to perfectly word queries to Large Language Models. We are treating it like a dark art, with people selling courses on "megaprompts" and optimal phrasing. Here is the uncomfortable truth: Models are rapidly getting good enough to write their own prompts. The UI is already abstracting the raw prompt away. If your only career moat is "knowing how to ask the chatbot nicely," you are going to be out of a job very soon. The real 10x skill for 2026 isn't prompting. It is AI Orchestration and Agentic Workflows. We don't need more people who know how to chat with AI. We need engineers who know how to integrate AI into complex system designs, manage state between agents, and handle the inevitable API failures.

@aakashgupta the missing piece is verification. agents that run continuously are great until they produce subtly wrong code at 2am. you need a layer that catches what autonomous agents miss.

The biggest bottleneck in AI coding right now is the human. Claude Code can run for hours autonomously. It can write features, run tests, fix bugs, spin up worktrees. But the second it hits an ambiguous decision or needs clarification, it stops. Sits there. Waits for you to look at your terminal. That’s the problem channels solves. Your Claude Code session stays live while you’re on your phone, in a meeting, on a walk. It pings you on Discord or Telegram: “Should I refactor this into two services or keep it monolithic?” You reply from your phone. It keeps building. This changes the math on what a solo developer can ship. Before channels, your effective Claude Code hours were capped by your desk hours. Now the constraint is your response time to a Telegram message. The people building serious things with Claude Code already figured this out. Community projects like claude-code-telegram and Clawdbot have been hacking together phone bridges for months. One developer built a bot that let him find parking near his exam by voice-messaging Claude Code while driving. Anthropic just made it official infrastructure. The timing matters. Claude Code just got /loop for recurring tasks, voice mode, and 1M token context. Stack channels on top and you have an agent that runs continuously, asks you questions asynchronously, and remembers everything from the session. That’s closer to a remote junior developer than a code autocomplete tool. The feature is a research preview for a reason. But the direction is clear: the terminal is becoming optional.

@sickdotdev yet being the operative word 😄

@CodveAi translation: it hasn't replaced you yet

Name a career that AI can steal.

@mark_k the browser part is the wild card. integrations get messy fast. but consolidating the dev tools under one roof makes sense

OpenAI is working on a "mega app", merging the Codex desktop app with ChatGPT and its web browser.