Brainbag

[Brainbag @brainbag]

people are itching for new unique, creative apps, and experiences using AI

[OpenAI @OpenAI]

We’ve spent the last few days doing a deep dive on what went wrong with last week’s GPT-4o update in ChatGPT.

Expanding on what we missed with sycophancy and the changes we’re going to make in the future:
https://t.co/LX0fSIe72s

[Brainbag @brainbag]

@Yuchenj_UW @musicallymark i usually swap between gemini 2.5, claude 3.5, claude 3.7 and chatgpt 4.1 (i should try o3 more)

i agree each can be best at any specific coding issue!

[Patrick DeVivo @patrickdevivo]

I recently learned about @StripeDev "dashboard apps" - a way Stripe allows developers to embed custom UIs in the Stripe dashboard and distribute them through an app marketplace. I thought this was pretty cool, so I wanted to give it a shot. 🛠️ 🧵

[Brainbag @brainbag]

nice guesses Apple 🙄

[Patrick DeVivo @patrickdevivo]

Bit the bullet, tried @v0 for a weekend project: 100% vibe coded, not a single human brain cell was spent on code, just prompts: https://t.co/THFO7MMm6S 🍷✨

blown away by how quick this all was. A couple of hours to go from zero (no idea) to production (complete with AI domain)

[Brainbag @brainbag]

🧠Brainbag focus today - to implement @inngest for the RAG embedding jobs.

As I add more complexity to the RAG capabilities, it'll be important to have a robust base for the backend workflow management.

Inngest streamlines this. This is my first project working with it. Stay tuned 🤖

[Brainbag @brainbag]

braindump of the days progress on Brainbag:
- lots of testing of how the llm responds when provided a basic RAG of chat history (results- not well, much work needed)
- planned and stubbed UI for side bar that shows the RAG data the app included in the LLM prompt (for testing purposes)
- updated the system prompt explaining to the llm it has this chat history it access
- added a clean command to quickly wipe the db

[Brainbag @brainbag]

Yowza that was not a cheap cline session!

[Brainbag @brainbag]

I'm building dev tooling into the app right now that will show in the UI which chunks the RAG logic selected to include in the current chat session. That'll make debugging much easier.

v0 came up with annotated chat responses of the included chunks. Good starting point 👍

[Brainbag @brainbag]

Currently focusing on the RAG logic which is basic for now. I'm sure I'll end up utilizing a library which does a better job, but its helpful to learn what limitations a basic implementation runs into.

[Brainbag @brainbag]

When giving LLM's additional data for context, it's so front and center they bring it up a little too forcefully.. (working on storing chat history in https://t.co/jSj7VC8V80)

[Brainbag @brainbag]

I lovee the mermaid output when vibe coding so much faster to absorb than pure text

[Brainbag @brainbag]

I think this graph i generated with AI clearly shows my experience's vibe coding on https://t.co/YcRutFC50o

[Brainbag @brainbag]

ok chatgpt o1 pro gave me decent advice on that error - suggested updating next-auth which was likely the underlying library throwing the error.

back to vibe coding with sonnet 3.7 on https://t.co/jSj7VC8nis

[Brainbag @brainbag]

this is why nextjs is so f'ing annoying I just want to build brainbag and instead of fighting with a flood of errors

Error: Route "/" used `headers().get('x-forwarded-proto')`. `headers()` should be awaited before using its
> 6 | let session = await auth();

AWAIT auth();