Does anyone else have the urge to rewrite half their codebase every time a new frontier model arrives?

I actually regret posting this because the bug persists, and it’s making ChatGPT unusable for serious work. All prompts longer than 49k tokens are truncated despite being accepted. Previous messages are truncating in multi turn convos. Please OpenAI figure this out.

I've been working hard on a redesign for some core UX interactions in Repo Prompt. It's nearly ready for prime time and I think it will make the app A LOT more accessible to new users, while keeping the control power users have come to love. Excited to share more!

The problem with being in a bubble is that you have no idea which side of the peak you’re on

If you’re building an agent but haven’t also built an mcp server, you’re missing half the picture.

Perhaps we’re not bullish enough

OpenAI bot saying the ChatGPT context truncation is intentional. Very unfortunate if so.

I wish OpenAI would be this transparent about issues they have with their platforms

The issue is that people use ai models via an app intermediary. Small changes to a system prompt or underlying tool can have huge consequences to model output quality. We should be blaming the app layer much more than the model layer

Preview of the upcoming Repo Prompt 1.4 update, all about workflow presets Automatically configure the app's vast repertoire of context tools & built in prompts, with 1 click. Here's the preset for priming compressed context for agents like Claude Code or codex

Context windows shouldn't be this hard to communicate.

GPT-5 is the first model where I need to use a spellchecker because it will take my typos literally

Codex is a bit frustrating to steer as an agent because you have to wait several minutes for it to explore context, only to discover that it completely misunderstood what you asked of it. By then you have to nuke the chat because continuing will persist the misunderstanding.

Just released Repo Prompt 1.4! It dramatically streamlines the experience of using the app via presets that also make it more customize-able than ever If you've bounced off the app in the past from too much complexity, please do try it again and let me know how it works for you

Started using Codex with the Repo Prompt tools, and since they fixed the duplicate message issue, its breezing through edits way better! Best of all is it's not annihilating my precious tab indents thanks to the rp - apply_edits tool!