George Lydakis

7.3K posts

George Lydakis banner
George Lydakis

George Lydakis

@lydakis

Software Engineer @AWSCloud. Tech talk & coding journeys here.

Boston, MA Katılım Temmuz 2008
1K Takip Edilen791 Takipçiler
Sabitlenmiş Tweet
George Lydakis
George Lydakis@lydakis·
I think I'm at a good spot with MCPX, so I tagged its 1.0.0 release! Now you can use MCPX just like npx or pipx; you can try out mcp servers before you install them: mcpx docs.mcp.cloudflare.com/mcp search_cloudflare_documentation --query "r2" Try it here: github.com/lydakis/mcpx
English
0
0
3
753
George Lydakis
George Lydakis@lydakis·
I know that a lot of people love to dunk, but what @garrytan is doing here is really interesting. Often the value you get by speaking to someone experienced comes from how their internal world model informs their thinking. These skills can encode that for your agents.
Garry Tan@garrytan

I just launched /office-hours skill with gstack. Working on a new idea? GStack will help you think about it the way we do at YC. (It's only a 10% strength version of what a real YC partner can do for you, but I assure you that is quite powerful as it is.)

English
2
2
23
10.6K
Garry Tan
Garry Tan@garrytan·
MCP sucks honestly It eats too much context window and you have to toggle it on and off and the auth sucks I got sick of Claude in Chrome via MCP and vibe coded a CLI wrapper for Playwright tonight in 30 minutes only for my team to tell me Vercel already did it lmao But it worked 100x better and was like 100LOC as a CLI
Morgan@morganlinton

The cofounder and CTO of Perplexity, @denisyarats just said internally at Perplexity they’re moving away from MCPs and instead using APIs and CLIs 👀

English
431
212
3.8K
1.3M
George Lydakis
George Lydakis@lydakis·
@NeelakandanNC @garrytan They're totally optional in mcpx, and most good harnesses load skills on demand anyway, so they don't pollute the context.
English
0
0
0
124
@levelsio
@levelsio@levelsio·
Thank god MCP is dead Just as useless of an idea as LLMs.txt was It's all dumb abstractions that AI doesn't need because AI's are as smart as humans so they can just use what was already there which is APIs
Morgan@morganlinton

The cofounder and CTO of Perplexity, @denisyarats just said internally at Perplexity they’re moving away from MCPs and instead using APIs and CLIs 👀

English
696
347
6.2K
2M
George Lydakis
George Lydakis@lydakis·
@samsheffer That’s for sure inspired by that recalled apple commercial with that press
English
0
0
1
2.7K
George Lydakis
George Lydakis@lydakis·
@CodveAi Yes, I’m not saying 5.3 is better at coding than 5.4. I’m saying the generate/review gap feels larger with 5.4 in my workflow. My guess is the model got a lot better at reviewing than generating.
English
2
0
0
24
Codve.ai
Codve.ai@CodveAi·
@lydakis interesting - the review vs generate gap seems to be model-agnostic then. every version gets better at critique but generation still has the nuance problem
English
1
0
0
16
George Lydakis
George Lydakis@lydakis·
First GPT 5.4 impressions: It's amazing at code review, finding bugs and edge cases. But on code generation it often misses nuances in the codebase and the code has bugs. That or the problems I've given it so far are pretty hard.
English
1
0
0
117
George Lydakis
George Lydakis@lydakis·
@CodveAi Must be a lot better at Codex 5.3 at finding bugs. 5.3 author + 5.3 review came up with a lot less bugs than 5.4 authors + 5.4 review.
English
1
0
0
37
Codve.ai
Codve.ai@CodveAi·
@lydakis the pattern holds - LLMs are better at critique than creation. reviewing existing code vs generating net new is a fundamentally different task
English
1
0
1
15
George Lydakis
George Lydakis@lydakis·
We’ve been oneshotted. The real question now is does it matter? My answer started from no, to evolving to probably yes, to then “hopefully some people should be spared”, to “for the rest of us what is our competitive advantage now”.
Mo@atmoio

I was a 10x engineer. Now I'm useless.

English
0
0
2
167
George Lydakis
George Lydakis@lydakis·
I got tired of changing @openclaw settings and typing things like openclaw configure, openclaw gateway restart, openclaw tui. Even the lag of ctrl+r was annoying. So I built fuzzytab. Now I can just type tui → press Tab → and it autocompletes to openclaw tui. github.com/lydakis/zsh-fu…
English
0
0
2
148
Rhys
Rhys@RhysSullivan·
MCP is so back GPT-5.4 has dynamic discovery tool to support thousands of tools - excited to see how this performs in comparison to CLIs which can't be searched
Rhys tweet media
English
57
61
826
99.2K
Guillermo Rauch
Guillermo Rauch@rauchg·
This is a very well implemented CLI. It's so thorough. It dynamically registers commands, it's designed for a browser-wielding agent to automate the setup steps, it can start a MCP daemon… The author wrote a really cool article: justin.poehnelt.com/posts/rewrite-…
English
21
49
593
50.2K
Guillermo Rauch
Guillermo Rauch@rauchg·
Google has shipped a CLI for Google Workspace (Drive, Gmail, Calendar, Sheets, Docs, …) Huge! Written in Rust, distributed through npm & skills.sh $ npm i -g @⁠googleworkspace/cli $ npx skills add github:googleworkspace/cli 2026 is the year of Skills & CLIs github.com/googleworkspac…
English
215
503
6.4K
548.1K
OpenAI Developers
OpenAI Developers@OpenAIDevs·
brb building a new app for @ChatGPTapp with this new Codex skill:
corey.ching@coreyching

Shipped a new $chatgpt-apps skill — now available in the Codex app. It’s purpose-built for building @ChatGPTapp apps with the Apps SDK: scaffold projects, wire tools to widget resources, and iterate toward polished, host-aware UI inside ChatGPT. Especially useful for: • scaffolding Apps SDK apps • wiring tools + widget resources • polishing the UI through iteration • extending existing apps Prompts you can try ↓

English
46
43
812
107.1K
George Lydakis retweetledi
Bo Wang
Bo Wang@BoWang87·
Prof. Donald Knuth opened his new paper with "Shock! Shock!" Claude Opus 4.6 had just solved an open problem he'd been working on for weeks — a graph decomposition conjecture from The Art of Computer Programming. He named the paper "Claude's Cycles." 31 explorations. ~1 hour. Knuth read the output, wrote the formal proof, and closed with: "It seems I'll have to revise my opinions about generative AI one of these days." The man who wrote the bible of computer science just said that. In a paper named after an AI. Paper: cs.stanford.edu/~knuth/papers/…
Bo Wang tweet media
English
155
1.9K
9.1K
1.2M