Your rules file: loaded, or only mentioned?
Claude Code skips AGENTS.md by default if a CLAUDE.md sits in the folder or above. /context shows what loaded.
Codex has no imports: "read docs/rules.md" in AGENTS.md is only a request.
pathrule doctor shows what each agent gets.
The file says "Do not edit." Your agent edits it anyway. Next regenerate, the fix is gone.
Same request with Pathrule in Claude Code: the edit is stopped before it lands, and the rule names itself. Then it changes the spec.
pathrule.io#ClaudeCode#Codex#DevTools#ai
"Stop. Show me the plan first." Typed Monday. Again Tuesday, in a new session.
With Pathrule, the second time it sticks: kept in your own words, never the agent's paraphrase. On Wednesday, Claude Code opens with the plan.
pathrule.io#ClaudeCode#Codex#AICoding
Claude Code hits its limit mid-refactor in apps/mobile. You open Codex.
Without: you re-explain that routing is Expo Router and Redux was dropped in May.
With Pathrule: Codex gets the same rules and decisions for apps/mobile before it writes. They never lived in either agent.
@lunkertw Exactly. This is one of the problems we’re addressing with Pathrule. Keeping memories isn’t enough. We also handle conflicting and outdated knowledge so agents don’t keep relying on decisions that no longer apply.
Your coding agent takes notes on you. Where they go:
Claude Code: ~/.claude/projects//memory/, never expire
Codex: ~/.codex/memories/, off by default
Copilot: on GitHub, deleted after 28 days unused
OpenCode: writes none
Seven agents, checked:
pathrule.io/writing/where-…
First day, in billing. She asks her agent for webhook retries. It stops her with what the team decided in March.
"How do you know that? I started today."
"So did I."
Pathrule: nobody starts from zero. Open source.
pathrule.io#AICoding#Onboarding#DevTools
One typo on a 404 page. Your agent loaded the payment rules, the iOS rules and the Kafka rules first. 94 rules, every turn. It needed two.
Pathrule gives your agent only the rules of the folder it works in. Open source.
pathrule.io#ClaudeCode#AICoding#DevTools
You tell your agent once: we dropped GraphQL, from now on new endpoints are REST under /v2.
Pathrule keeps that sentence as a decision, in your own words. Next session the agent already knows.
pathrule.io#ClaudeCode#AIcoding#DevTools
Your AI can read the code. But can it explain why your project works this way?
Pathrule Intelligence turns the decisions, constraints and knowledge behind your project into intelligence your agents can actually use.
The agent can change. The intelligence stays.
#ai#claudecode#codex#developer
Pathrule doesn’t treat an agent’s conclusion as a permanent truth by default. Clear, directly verifiable learnings can become memories, while useful but more subjective ones are surfaced to the user as suggestions. Each run also leaves a structured trail in the Activity Log: what happened, where it happened, what knowledge was used, and how it turned out. So durable context isn’t built on a single agent’s judgment, but on source, scope, actual work history, and user control when needed.
Stop briefing your AI about the same project over and over again.
Pathrule remembers how your projects work and uses a local LLM to keep learning as you build, so you can spend less time rebuilding context and more time focused on what you’re actually trying to achieve.
#AICoding #ClaudeCode#Codex#DevTools#AI
Every coding agent starts over.
Pathrule makes sure what they learn doesn't.
Explain it once. Let every agent build on it.
pathrule.io
p.s. the entire video was made with Opus 5.5 👀
AI agents are getting better. Context is still the problem.
Pathrule gives every agent the right memories, rules, skills, and signals for the code they’re working on, then carries what they learn back into your product context.
Now that loop goes further.
Design stays connected through Figma. Changes become planned work. Agents keep it moving.
One product. Every agent in sync.
pathrule.io
@aaliya_va Thank you so much, Aaliya! You’ve been supporting Pathrule since the very beginning, and we truly appreciate it. Glad Signals caught your attention. A lot more is coming! 🚀
4K Followers 1K FollowingBuilding 30 working mini web apps in 30 days. Starts Nov 1. Follow the daily launches → https://t.co/lPa6jgIWuW. AI agents, apps & automations.
40 Followers 2K FollowingChecking out the latest tech and business news that I'm excited about and hopefully something will be awesome to see and check out.
12K Followers 1K FollowingLearn AI Today and Future proof yourself. I help you to 10x Your Business with AI. 📩 DM or Mail for collab: [email protected]
41K Followers 5K FollowingAI l Researcher | Finding and sharing Al tools every day | Building tomorrow's tech | Open to collaborations
💌 [email protected]
18K Followers 6K FollowingTesting Al tools, exploring trends sharing what works For Collab DM for paid collaboration
📨[email protected]
https://t.co/78UOJDhwue
17 Followers 130 Following🎮 QA para videojuegos indie | 🔍 Encuentro lo que rompe tu juego antes de que lo vean tus jugadores | 📢 Comparto lo que aprendo testeando proyectos reales
2.0M Followers 2 FollowingClaude is an AI assistant built by @anthropicai to be safe, accurate, and secure. Talk to Claude on https://t.co/ZhTwG8d1e5 or download the app.
600K Followers 53 FollowingThe Gemini app turns research into reality, bringing frontier AI experiences like Omni, Nano Banana, and more to over a billion people.
5.5M Followers 4 FollowingOpenAI’s mission is to ensure that artificial general intelligence benefits all of humanity. We’re hiring: https://t.co/dJGr6LgzPA
1.8M Followers 2 FollowingWe're an AI safety and research company that builds reliable, interpretable, and steerable AI systems. Talk to our AI assistant @claudeai on https://t.co/FhDI3KQh0n.