Donnie Flood

Context engineering geek. 👋 CEO @ContextBridgeAI. Former LinkedIn, Homebound, GuildEducation, Bizo, etc. Remote Cameron Crazy (Go Duke!)

[Donnie Flood @floodfx]

@prem_k "measure the shoreline" is a great way to put it!

[Donnie Flood @floodfx]

@stockthoughts81 Ha. Thats what we do with your markdown plans so you can actually refine it like a human: https://t.co/gOF5bJrBhi

[Donnie Flood @floodfx]

@hunvreus Giving feedback on a markdown file in a terminal also sucks! Are you using a tool to remove the friction around plan refinement?

[Donnie Flood @floodfx]

"I’m considering whether I need to read the plan I wrote. I thought maybe I don’t need to"
-- GPT 5.5 in a reasoning trace
🤣

even the AI doesn't want to read AI generated text!

#clankerfail

[Donnie Flood @floodfx]

I assume you don't just approve the plan but refine it first? Using any tools to help with that? PlanBridge is open source and built specifically to make it easy to select text, add comments, and easily go from first draft of a plan to great one. Check it out: https://t.co/AlngUYZgVx

[Donnie Flood @floodfx]

"Load-bearing" is the new "you're absolutely right"

[Donnie Flood @floodfx]

Played a bit with Antigravity with Gemini 3.5 Flash. Damn is it fast! Like blazing! 🔥🔥🔥

[Donnie Flood @floodfx]

@InduTripat82427 Wouldn't it be nice if you could select text in the rendered markdown plan, add comments, and easily iterate on the spec? Here's an open source tool that does just that: https://t.co/vQljdhBaeD

[Donnie Flood @floodfx]

@eliheuer the 🔥 is 🔥

[Donnie Flood @floodfx]

@AntonMartyniuk Giving feedback to agents in "plan mode" sucks. You should recommend a tool like PlanBridge to help people not settle for the first plan Claude generates. https://t.co/vQljdhBaeD

[Donnie Flood @floodfx]

@DivyanshT91162 Looks like it is still outputting plans in raw markdown? Could be good to use a tool like PlanBridge to give precision feedback to the agent on the markdown it creates? https://t.co/vQljdhBaeD

[Donnie Flood @floodfx]

@cyrilXBT love me some "plan mode" and this open source tool makes it so much easier to give precision feedback and iterate on a plan. Try and and LMK what you think:
https://t.co/vQljdhBaeD

(disclaimer - I am one of the authors)

[Donnie Flood @floodfx]

If you've ever settled for a less than ideal plan from a coding agent because iterating in a terminal sucks you should try PlanBridge: https://t.co/gOF5bJs96Q

[Donnie Flood @floodfx]

@AnalyticsVidhya Nice. We have a GitHub app that keeps your agent context files updated automatically. We’re looking for early feedback if you’re up for trying it: https://t.co/7uGoTfkFy5

[Donnie Flood @floodfx]

@aidenybai I assume this reviews your local sessions? How does it dedupe and decide what goes into your CLAUDE.md?

[Donnie Flood @floodfx]

@ds_serena_ I'd be curious to see the 24 rules it came up with... Would you share?

[Donnie Flood @floodfx]

@VibeCoderOfek Nice! How are you keeping CLAUDE.md up to date as your codebase evolves?

[Donnie Flood @floodfx]

@vishisinghal_ Curious how you're keeping CLAUDE.md from rotting over time as your codebase changes?

[Donnie Flood @floodfx]

ContextBot is free during our public beta. Install the GitHub App and get your first context-improvement PR in minutes.

ContextBot stops the context rot.

https://t.co/EBkUQOBP5Z

[Donnie Flood @floodfx]

Your CLAUDE.md used to be accurate. 200 PRs later? It's full of legacy instructions. And your agent is a confused slop-machine.

[Donnie Flood @floodfx]

ContextBot does more than read PR comments. It learns from the code you merge every week. The diff itself carries signal about how your team writes code. That's where the implicit patterns live.