The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
Most AI builds break at the wiring, not the model: no retries, no log of what the model actually did, and no clear line for what the agent may do without a human.
I build that layer. My own products run on it -- an ordering platform where an AI agent handles real orders, and a coding agent that writes no file until someone approves the change.
The question worth answering first is what your agent should never decide on its own.
For the coding agent, I'd bind approval to both the proposed diff and the file's current version. If another edit lands between review and write, does it stop and request a fresh review?
Great question Kapil. Approval tied to the exact change is the core of how I build it, so nothing gets written that a human did not see first.
Agreed. Money is the clearest line there is. If it can't be undone, a human signs off.
For me the line is anything irreversible or anything that costs money: refunds, deleting data, messaging customers in bulk. Below that line the agent can act, as long as every tool call is logged with its input and result, so you can replay what it actually did. The detail I'd...
Well said Mojtaba. Logging every tool call with its input and result is what lets you trust the agent. You can always see exactly what it did.
Exactly. The one extra thing I log is the decision itself, not just the call: why a tool was chosen or skipped, and whether a human was required. When something breaks, "what it did" plus "why it decided to" is what makes the run replayable instead of a guess. Are you storing...
Agreed Mojtaba. Logging the reason behind each tool choice turns the log into something you can actually learn from.
This is exactly right — the model is rarely the problem, the wiring is. Approve-before-any-write on the coding agent is the detail that matters. A file write looks small until it's the wrong file. My line building agentic systems is similar: anything irreversible, or anything a...
Thanks Olanrewaju. Exactly, a file write looks harmless until it is the wrong file. Good to see you draw the same line.
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
An exploration of art, creativity, and creative direction through the lens of AI 🌸
This is an independent, experimental concept project, created entirely with AI to explore new possibilities in visual storytelling, artistic expression, and creative direction.
The work incorporates AESOP brand assets and identity, which remain the property of their respective owners. This is a personal creative exploration, not an official AESOP campaign or commissioned work.
Created with AI. Directed by me.
A concept. A creative experiment. A vision brought to life through AI.
So cool!!
Excited to share the latest web redesign projects for Codex⚡
The goal was to pair technical credibility with a more approachable, modern financial aesthetic while preserving the depth and complexity that make Codex a high-profile data provider.
The result is a visual direction designed to resonate across the spectrum, from crypto builders and neofinance founders to enterprise-scale exchanges like Coinbase. 💻
Love the logo on the keyboard key