---
title: "Minh-Toan Ta · Consonance"
description: "Minh-Toan Ta, full-stack engineer in Valence, France. Projects, articles and tools for AI agents — ideas that resonate together to shine brighter."
canonical: "https://me.consonance.fyi/en/"
lang: "en"
---

# When voices resonate, the night lights up.

> Consonance is the moment different sounds vibrate together and grow stronger. That is how I build: systems, ideas and people that resonate together, to shine brighter.

_con·so·nance_ /ˈkɒn.sə.nəns/ — from Latin consonare, “to sound together”

## What I’m building

- [Consonance](https://me.consonance.fyi/en/projects/consonance.md): A daily radar for AI and agents.
- [tracé.](https://me.consonance.fyi/en/projects/trace.md): Public transport traffic in France, network by network.
- [Luminode](https://me.consonance.fyi/en/projects/luminode.md): One body of knowledge. Paths that fit you.

## Latest articles

- [Grok Bot: what I would trust an always-on agent with](https://me.consonance.fyi/en/blog/grok-bot.md): Grok Bot, seen by an engineer who builds systems that must not be wrong: how xAI's agents work, what real marketplace Bots do, and my rules before I let one in.
- [OSS Scanner: a free security inspection for critical open source](https://me.consonance.fyi/en/blog/anthropic-oss-scanner.md): Anthropic's OSS Scanner runs periodic AI security scans of critical open-source projects at no cost. What a report contains, who qualifies, how to enroll, and how to triage the results.
- [pstack: Lauren Tan's rules for agents you can trust](https://me.consonance.fyi/en/blog/pstack-lauren-tan.md): pstack, the skill plugin by Lauren Tan (@poteto) for Cursor and Grok Bot: trust one agent before running ten, fix mistakes at the root, and treat verification as infrastructure, down to TLA+ models.
- [Claude Code in the cloud: one machine per task](https://me.consonance.fyi/en/blog/claude-code-in-the-cloud.md): A cloud session gives each task its own machine and branch: several jobs run side by side, survive a sleeping laptop, and come back as pull requests.
- [Evaluating and improving an AI app, explained with an exam](https://me.consonance.fyi/en/blog/automating-eval-design-and-hillclimbing.md): build-eval writes an exam for your AI app with you; hillclimb revises against it one change at a time, with held-out cases to catch cramming.
- [Moving to Claude Sonnet 5.5: the three things that break](https://me.consonance.fyi/en/blog/building-with-claude-sonnet-5-5.md): Sonnet 5.5 is 30% faster than Sonnet 5 at the same per-token price. Swapping the model name is easy; three habits in your code are not.
- [Claude Code mods: a small file that rewrites what you see](https://me.consonance.fyi/en/blog/getting-started-with-claude-code-mods.md): What a Claude Code mod is, how it slots into the session as a hook, two examples from the official guide, and Recording Mode to hide secrets on screen.
- [Claude Code's effort, explained: spending time, not intelligence](https://me.consonance.fyi/en/blog/spending-your-effort.md): Effort doesn't make Claude smarter. It sets how much computing it spends, so how much it checks, tests edge cases and judges on its own.

## Full-stack engineer, night-time explorer

Since 2022 at Conduent, I build ticketing and bank-card payment servers deployed across 20+ cities. Alongside, I explore AI agents.

---

- HTML: https://me.consonance.fyi/en/
- [Français](https://me.consonance.fyi/index.md) · [Tiếng Việt](https://me.consonance.fyi/vi/index.md)
- MCP: https://mcp.consonance.fyi
- Contact: tmtoan2601@gmail.com · https://www.linkedin.com/in/minhtoanta/