Krysoph

@viktoroddy As someone who loves to read, I just hate those websites full of animations with little to no information about what it even means.

Google Antigravity is insane. ❤️‍🔥Just recorded a 19-min tutorial on how to build this animated, award-winning website with Antigravity + GPT Image 2!

@setyamickala @rtsitola @XiaomiMiMo I've spent 4 billion tokens in DeepSeek in half a month, cost me $40. 82B seems so cheap for $88.

@rtsitola @XiaomiMiMo 1 month

Free Api Key @XiaomiMiMo Region: China Api key: tp-cap7t1dl21hzdl3a28kmmacmptiemfczt5rm81ekd8xnku6k Base URL OpenAI: token-plan-cn.xiaomimimo.com/v1 Base URL Anthropic token-plan-cn.xiaomimimo.com/anthropic Model : mimo-v2.5-pro

I had an idea about a month ago about how S3 could be used as server for git if everything was done in the client side, there is only one issue: Authentication. I went as far as building an git alternative called got that does everything client side using If-Match to avoid races but I haven't figured out the auth part yet other than requiring a server. The website would be statically generated on every push and anything other than the repository page would load a DuckDB database through WASM and have all the info like commits, issues, pull requests, contributors and file metadata instantly loaded as you navigate. Then we could have a real GIT HUB to connect all the external S3 repos into a single website that only serves as a moderated central database but has no power over any repo other than showing it exists and can have it's data publicly accessed.

I'm going to use my AI psychosis to fix clouds for agents. Someone else needs to use their psychosis to fix source control. I would do it myself but I'm already too deep on the cloud thing. GitHub is dying and git is not the right primitive. Will dump some thoughts here.

Found Qt6's real weakness while building my native AI chat: EMOJIS. They're shockingly heavier than anything else: - Tables - Syntax highlighted code blocks - Markdown - HTML Anyone else hit this with QWidgets? How did you solve it?

@neerajjj6785 MiMo v2.5 Pro has the best price/perf at the moment.

POV : The AI race in the last 5 years which AI is leading right now?

DeepSeek and Xiaomi-MiMo V2.5 now share the same price in both Flash and Pro models. Seems fair, considering that both have about the same score on arena.ai/leaderboard/te…

@ThePrimeagen Grandma likes that style... on her.

I think I have too much hair

@claudeai You forgot Bun's Zig to Rust

Six Claude projects that all came from the same question: “why not?”

@DoppioJP @dhh He did something that locks the font size, can't change it from the browser's zoom.

@Krysoph @dhh Cmd/ctrl - and you have smaller fonts if you want. Interestingly that works on Safari but not on Chromium browsers. 🤔

Basecamp 5 is live! And the new website is a throwback to a simpler time. Just show the damn product. Just list the features. Yes, it's agent accessible, but it's not agent hysteric. basecamp.com

@shl 🤣Specially when you want to play like Allan Holdsworth.

Every software engineer is a failed musician

AntiGravity is just the worst at giving any kind of information about quota. Not even the reset time is displayed anymore.

I'm building a native AI chat in C++/Qt6, it's called Zeteo. It uses about 30-40MB ram, no GPU and during heavy tasks up to 10% CPU (in a Ryzen 3600). Theres some harness: memory (with search), file related tools (read/write files, walk dirs). Current features: file editor with syntax highlighting, project directory tree, project tabs, editor layout switch (choose editor on left or right), file upload, custom themes, DeepSeek and OpenRouter support. Theres still a lot to do, editor diff, git integration, improve ram usage, add support for the major APIs (OpenAI, Anthropic, Grok...), DB sync to S3...

looking to connect people on X if you're into - building SaaS - vibe coding - AI tools - shipping in public - figuring it out as you go say Hi or drop what you're working on looking to follow active ones 👋

@LexnLin Looks nice, I'll take that as inspiration to improve my native AI chat.

grok build is cool

@Im_IrushiK Let it keep working as a quarantine.

Unpopular opinion: Elon Musk should buy LinkedIn and delete it

Works well for implementing fast but it fails on planning when compared to GPT 5.5, and it loses efficiency on very long chats getting kinda dumb. It's way ahead Gemini 3.5 Flash in "smartness". I wish I could afford to keep it as my main model, when the quota run out I'll have to rely on DeepSeek, which not a Cursor problem btw. I'd also like to implement the SDK in my native AI chat (Zeteo: C++/Qt6), is there any plan for a WASM version of the SDK?

what do you think of composer 2.5 so far? how can we make the next model even better? want to hear your feedback on behavior, speed, quality, whatever!

@shadowe1ite @damnSuryansh @BadalK99277 try shrek

@damnSuryansh @Krysoph @BadalK99277 Try superman

Can you notice the difference when you search Linux on DuckDuckGo?

@Rushpcs All good, just make sure to use Medium reasoning level most of the time instead of Extra High to avoid wasting resources. I use: - 5.5 Extra High for planning - 5.3 Medium for implementing - 5.4 Medium for difficult tasks

@Krysoph Already baught codex because of the hype i will invest in this next month

Guys i have 20$ Should i get codex ? I have claude code too!

@elonmusk You could connect New York to Greenland by making a tunnel below the ocean like in France-UK.

Boring Company’s goal is to solve traffic in cities and one day build Hyperloop high-speed transit between city centers