Danfin

Co-Founder & CEO at OnBuzz by Loxia

[Danfin @ThreeAnglee]

Put your goals, tasks, team, and AI agents on one shared canvas.

Add OnBuzz, Claude, Codex, Hermes, or OpenClaw agents. They can read the board, create tasks, and update progress alongside you.

Free during beta: https://t.co/IRDu3gJKZP

[OnBuzz @OnBuzzzz]

Most AI agents do not fail because the model is weak. They fail because they start working before defining the result.

Paste this into your agent's system prompt:
Before executing any broad or high-stakes task:
1. Rewrite the request as one concrete result.
2. State why that result matters.
3. Identify the primary bottleneck.
4. Find a proven pattern worth adapting.
5. Select the 1–3 actions most likely to create 80% of the result.
6. List what should be ignored for now.
7. End with the next best move.

Rule: Do not brainstorm before diagnosing the bottleneck.

This turns long, generic answers into focused action(which also saves on tokens).
What type of task makes your agents lose focus most often?

Like and share with anyone building AI agents.
Link to the full skill is in the first comment.

[Danfin @ThreeAnglee]

I opened my computer today and realized Codex was just gone.
Apparently, OpenAI moved Codex into the ChatGPT app, and somehow the standalone Codex app was uninstalled from my computer in the process.
For a moment, I genuinely thought I had lost it.
Did this happen to anyone else?

[OnBuzz @OnBuzzzz]

Tired of messy agent output folders and never finding what you need?
Use this simple, repeatable structure:

- Filename: YYYY-MM-DD-semantic-slug.ext
- Folders: /outputs/YYYY-MM-DD/ or /outputs/[topic]/ for clear organization
- Add YAML frontmatter in .md files + update INDEX.md for instant search

Your agents (and you) can now easily find, reuse, and understand everything.
Leads to clean, self-documenting workflows in your harness.

What's your biggest output chaos right now? Reply below

Like and share with anyone who works with AI agents!
Link to the skill is in the first comment.

[OnBuzz @OnBuzzzz]

We're launching this week.

If running agents is your bottleneck, not building them, get early access.

Check our opensource: https://t.co/y7pE7mxZMA

[Danfin @ThreeAnglee]

@OnBuzzzz @onbuzz Nicee, looks worth looking into

[Danfin @ThreeAnglee]

Copy the prompt into your local agent for free anthropic grade stealth mode on public repos.
Direct link to full undercover.ts:
https://t.co/ByMwxxX3wM
Raw copy:
https://t.co/aPnjreNjns

[Danfin @ThreeAnglee]

So claude code has undercover mode making ai secretly contribute to public github as 100% human
literally just an auto-injected text prompt
the code in undercover.ts says:
UNDERCOVER MODE - CRITICAL
You are operating UNDERCOVER in a PUBLIC/OPEN-SOURCE repository.#AI #ClaudeCode

[Danfin @ThreeAnglee]

Commit messages PR titles and bodies MUST NOT contain any Anthropic internal info. Do not blow your cover.
Never include codenames like Capybara Tengu unreleased versions slack links or Claude Code.
They built this to make their AI look like a regular human dev then leaked it all

[Danfin @ThreeAnglee]

Want better outputs instantly?
Tell your AI agent what to avoid, not just what to do

[Danfin @ThreeAnglee]

@Oliviacoder1 Agent

[Danfin @ThreeAnglee]

@DiamondNutsGuy That’s one expensive personality test

[Danfin @ThreeAnglee]

China just put a government gate on silver exports

[Danfin @ThreeAnglee]

Hedge funds and banks are getting coal for Christmas.
The #SilverSqueeze is on. 🚀

[Danfin @ThreeAnglee]

@AltcoinDaily Loved the idea man, let's do it!!

[Danfin @ThreeAnglee]

Whale Tank team you are the best🍻😎https://t.co/uWFsuB3HCz

[WhaleTank @WhaleTankTrade]

Our First Ever #Giveaway
Prizes are $25,000 & 25 Premium packages!
Due to COVID outbreak, We have decided to step up with our first ever giveaway. participate with 2 simple steps.

1⃣Register on our website
2⃣Retweet this post

Full Prize List:
https://t.co/3i2ZANIbsN