@pro.glitch: How to make multiple AI Assistant Part 2 #aiassitant #aiagent #professorglitch

Professor Glitch
Professor Glitch
Open In TikTok:
Region: US
Monday 18 May 2026 14:55:00 GMT
3163
105
5
44

Music

Download

Comments

ethiopiawiw_betarikachen
ቴዎድሮስ :
Man, you must be rich! I couldn't afford to by Cluaude's LLM model let alone to pay for 5 different Cluade accounts.
2026-05-18 18:01:10
0
chitownrobb
Robb :
I use Claude pro - I thought effectively. And I use Claude for my Hermes agent. I’d love to learn a bonified “system” do you have a video incorporating Hermes in this ?? Thanks.
2026-05-18 15:46:44
0
junito_ponce
Pedro Olavarria :
🔥👊
2026-05-19 23:22:46
1
aaronstarr1
Aaron Starr :
😳😳😳
2026-06-10 00:24:54
0
To see more videos from user @pro.glitch, please go to the Tikwm homepage.

Other Videos

You asked an AI to rename one function, and it left two files broken. That happens because most coding agents rename by searching for text: they find the obvious places, and miss the import, the re-export, the alias. oh my pi (omp) does something different, and it is the second idea worth stealing from this repo. It wires LSP into the agent as a tool. LSP, the Language Server Protocol, is the standard machinery your editor already uses to give you autocomplete, go to definition and error squiggles. It is the compiler answering questions about your code. Every serious editor speaks it. Almost no agent uses it. In omp the agent queries it directly instead of rereading your files: 14 dedicated operations, plus tools that work on the AST, the syntax tree of the code, so searches and rewrites understand structure instead of matching characters. What that looks like in practice, from their own capture: asked for the references of formatBytes, the language server answers with five references across three files. The rename then goes through the editor protocol, applying one edit in format.ts, two in report.ts, two in cli.ts. The plain text search afterwards returns zero matches, which is the proof nothing was left behind. The idea you can take today: let your agent ask the tool that already knows, instead of reading the file again. The compiler is never guessing about your code, and the model always is. Repo: github.com/can1357/oh-my-pi Install: curl -fsSL https://omp.sh/install | sh Follow for ai agents explained without the hype. #AI #aiskill #opensourceai #opensourceproject #ohmypi #github #claude #openai #opensource
You asked an AI to rename one function, and it left two files broken. That happens because most coding agents rename by searching for text: they find the obvious places, and miss the import, the re-export, the alias. oh my pi (omp) does something different, and it is the second idea worth stealing from this repo. It wires LSP into the agent as a tool. LSP, the Language Server Protocol, is the standard machinery your editor already uses to give you autocomplete, go to definition and error squiggles. It is the compiler answering questions about your code. Every serious editor speaks it. Almost no agent uses it. In omp the agent queries it directly instead of rereading your files: 14 dedicated operations, plus tools that work on the AST, the syntax tree of the code, so searches and rewrites understand structure instead of matching characters. What that looks like in practice, from their own capture: asked for the references of formatBytes, the language server answers with five references across three files. The rename then goes through the editor protocol, applying one edit in format.ts, two in report.ts, two in cli.ts. The plain text search afterwards returns zero matches, which is the proof nothing was left behind. The idea you can take today: let your agent ask the tool that already knows, instead of reading the file again. The compiler is never guessing about your code, and the model always is. Repo: github.com/can1357/oh-my-pi Install: curl -fsSL https://omp.sh/install | sh Follow for ai agents explained without the hype. #AI #aiskill #opensourceai #opensourceproject #ohmypi #github #claude #openai #opensource

About