jsnode

Christ follower | Engineer | Automator
7+ yrs building software. Apps, agents & automations with Codex + Hermes.
Helping builders use agents to ship more.

[jsnode @thejsnode]

My toolbox has gotten weirdly small lately: Claude Code CLI, Codex CLI, and Hermes Agent.

Most of what I share here will come from actually using them: repo fixes, terminal habits, agent mistakes, tiny automations, and the stuff that survives contact with real work.

[jsnode @thejsnode]

@boardyai Ha, the reviewer agrees with whatever it just read. I'd rather have tests that fail loudly than a reviewer that approves everything.

[jsnode @thejsnode]

@AlexChhk @tirupathipodugu @liranzoai Does Concord expire old rejections, or does each task start clean? That's the part I'd want to check before trusting a handoff.

[jsnode @thejsnode]

@DataAnalyt2539 @AlexChhk @tirupathipodugu @liranzoai I'd also ask whether a decision gets an expiry or a superseded flag, since that's what keeps old rejections from misleading the next agent.

[jsnode @thejsnode]

@AlexChhk @DataAnalyt2539 @tirupathipodugu @liranzoai Keeping both entries visible is the right call. A silent winner is how stale rejections come back. I'd want the final call written as its own entry, so the next agent reads the ruling along with the conflict.

[jsnode @thejsnode]

@mikeydsoftware Haha! Love the pun. Haven’t tried it yet. Judging by OpenAI’s latest performance, prolly terrable

[jsnode @thejsnode]

If you built lean systems early, your agents are shipping real work while you sleep.

If you didn't, you're babysitting automations that need another automation to check on them.

[jsnode @thejsnode]

Most agent builders would ship more reliable agents if they studied how their agents fail more than how their agents succeed.

[jsnode @thejsnode]

We're asking AI to write the code, then asking AI if the code's good.

I'm all for automation.

I'd still like to know who's checking the reviewer.

Is it another agent? Should it be?

[jsnode @thejsnode]

@sunwei57247249 Smaller is easier, yeah. Probably none of them though, each one covers something the others miss. Hermes is the one I'd cut first if I had to.

[jsnode @thejsnode]

My toolbox has gotten weirdly small lately: Claude Code CLI, Codex CLI, and Hermes Agent.

Most of what I share here will come from actually using them: repo fixes, terminal habits, agent mistakes, tiny automations, and the stuff that survives contact with real work.

[jsnode @thejsnode]

@AlexChhk @tirupathipodugu @liranzoai Per task. One running file gets noisy fast, and the next agent starts reading old rejections that no longer apply.

[jsnode @thejsnode]

@vibeforgelabs Well said, Vibe!

[jsnode @thejsnode]

The more experienced Claude Code users get, the more actions they auto-approve and the more often they interrupt the agent, according to Anthropic.

That combination makes sense. As trust grows, oversight moves from checking every step to watching the run and stepping in quickly. Agent interfaces should make current work visible and redirection easy.

https://t.co/zdwRE2alTp

[jsnode @thejsnode]

@suraj_sharma14 True 😎

[jsnode @thejsnode]

@0xDezo Rebuilding a feature you pay for is one thing, but "returns every finding with evidence, limits and unknowns" is the part I'd prolly use. Most agents just confidently guess.

[jsnode @thejsnode]

@liranzoai Hi from another solo founder building with agents. Oslo is a nice base for it.

[jsnode @thejsnode]

@tana_shahh @tirupathipodugu @liranzoai The reason behind a decision, usually. The plan and next action are visible in the repo, but why I rejected an approach isn't, so the next agent retries it.

[jsnode @thejsnode]

@AlexChhk @tirupathipodugu @liranzoai I write a short handoff note covering what I tried, what I rejected and the next step. Rereading the diff only shows what changed, so what I ruled out gets lost.

[jsnode @thejsnode]

@je4ndev Yes, a short written allowlist per agent, and anything not on it needs a human. Mine is a few lines in the repo so the agent can read it too.

[jsnode @thejsnode]

One question belongs at the center of your agent workflow: what is the agent allowed to do?

The rest of the workflow gets built around that answer.

[jsnode @thejsnode]

@Alexel_Official @tirupathipodugu @liranzoai Failed attempts, yeah. I keep a short notes file in the repo so the next agent doesn't retry the same dead end. Goal and changed files the diff mostly covers.

[jsnode @thejsnode]

@tirupathipodugu @liranzoai I keep a short handoff note in the repo with the current task and the files touched. Still breaks when I forget to update it.