RedEagle | Dave

@AtharvaXDevs My codex is stupid af, it’s insane how slow + dumb gpt is. I literally can’t work with that. Claude just does what I want it to do, gpt won’t.

NEVER BUY A CLAUDE CODE 20$ PLAN IT’S JUST A SCAM😭 NEVER BUY A CLAUDE CODE 20$ PLAN IT’S JUST A SCAM😭 NEVER BUY A CLAUDE CODE 20$ PLAN IT’S JUST A SCAM😭 NEVER BUY A CLAUDE CODE 20$ PLAN IT’S JUST A SCAM😭 NEVER BUY A CLAUDE CODE 20$ PLAN IT’S JUST A SCAM😭

No Claude, the project will not take me 2-3 months. We will finish it today.

@saltyAom Eden Query pretty pls ☹️

About next Elysia version - typebox 1 - much better type soundness for error handling - refactor, a lot of refactor, especially memory usage - make file-type optional - some undisclosed standard things - super secret thing

ModLog just processed its first 100,000 mod actions. Every timeout. Every ban. Every deleted message. Tracked, stored, visualized. Building the analytics layer that Twitch chat moderation never had.

@saltyAom You don’t have to justify yourself for anything. I’m shocked how many crazy replies are here. Do whatever you are comfortable with, don’t care about others. But I believe you already know this.

1. I’m not gay 2. I’m not on HRT 3. I’m a straight guy 4. I don’t have an OF

@saltyAom very excited about it 🔥

Eden Query

Never deleting this app🤣

Without telling me your age. What is the first video game you played? GIFS ONLY!!!

@mrpancakes39 @NeilTheFisher @saltyAom I just contributed to that project and it still has some smaller missing pieces, but for the most cases it will work. Indeed it’s not battle tested yet.

@NeilTheFisher @saltyAom But isn't too new? Can I use it in production safely?

@karpathy The only thing AI gave me was the satisfaction of my addiction to ship tools fast. But somehow that didn’t make me happy.

@karpathy I feel like I forgot everything about coding manually, without AI. I just canceled my LLM subscriptions and I’m going back to learn new frameworks and libraries by reading documentation and google when I don’t know stuff. I forgot that I actually love to code.

It is hard to communicate how much programming has changed due to AI in the last 2 months: not gradually and over time in the "progress as usual" way, but specifically this last December. There are a number of asterisks but imo coding agents basically didn’t work before December and basically work since - the models have significantly higher quality, long-term coherence and tenacity and they can power through large and long tasks, well past enough that it is extremely disruptive to the default programming workflow. Just to give an example, over the weekend I was building a local video analysis dashboard for the cameras of my home so I wrote: “Here is the local IP and username/password of my DGX Spark. Log in, set up ssh keys, set up vLLM, download and bench Qwen3-VL, set up a server endpoint to inference videos, a basic web ui dashboard, test everything, set it up with systemd, record memory notes for yourself and write up a markdown report for me”. The agent went off for ~30 minutes, ran into multiple issues, researched solutions online, resolved them one by one, wrote the code, tested it, debugged it, set up the services, and came back with the report and it was just done. I didn’t touch anything. All of this could easily have been a weekend project just 3 months ago but today it’s something you kick off and forget about for 30 minutes. As a result, programming is becoming unrecognizable. You’re not typing computer code into an editor like the way things were since computers were invented, that era is over. You're spinning up AI agents, giving them tasks *in English* and managing and reviewing their work in parallel. The biggest prize is in figuring out how you can keep ascending the layers of abstraction to set up long-running orchestrator Claws with all of the right tools, memory and instructions that productively manage multiple parallel Code instances for you. The leverage achievable via top tier "agentic engineering" feels very high right now. It’s not perfect, it needs high-level direction, judgement, taste, oversight, iteration and hints and ideas. It works a lot better in some scenarios than others (e.g. especially for tasks that are well-specified and where you can verify/test functionality). The key is to build intuition to decompose the task just right to hand off the parts that work and help out around the edges. But imo, this is nowhere near "business as usual" time in software.

@theo @__cruxaustralis In germany we say: Schlafen gelegt 😴

Going off memory, gonna miss a lot of things At Twitch, I built and shipped: - Twitch's internal and external moderation tools (mod view is one I'm really proud of) - The code that ran the Bob Ross Marathon - A bunch of random enhancements to Twitch Chat At T3 Tools (aka T3 Chat), we have built & shipped: - UploadThing (the best way to add file uploading to your apps) - Ping (the best way to bring a guest into your live stream) - MarkerThing (an essential tool for live creators chopping videos) - PicThing (thumbnail asset management and bg removal) - T3 Canvas (the new ai image gen experience, merged into t3 chat) - T3 Code (coming very soon) Personally, I have built & shipped: - The T3 Stack - Yerba (a popular Electron starter still used today) - The Stripe Recommendations guide that over 5,000 businesses have used to add payments - Unduck (search engine) - QuickPic (image conversion and editing tool) - SnitchBench (popular ai bench) I have contributed to dozens of open source tools since I started coding in high school. By the time I finished college, I had two apps with millions of downloads (InSquare for Pebble and Chrometana for Chrome/Windows 10). What have you built?

It’s really hard for me to respect a dev if they never ship anything

@steipete @jaayjayAT Du kannst verdammt Stolz auf dich sein und du bist ein Vorbild für mich. Alles richtig gemacht 🍀

@jaayjayAT Den "luxus" hab ich mir in 13 jahren grind erarbeitet. Den hatte ich davor auch nicht.

Nicht jeder hat den Luxus sich dann mal 2-3 Jahre Zeit für Burnout nehmen zu können. Danach mit einer 6-7 Tage Woche um die Ecke kommen wenn es wieder gut lauft. Naja. Ist ein Take, aber halt kein sonderlich guter für 99,9999% der AN. Viel Spaß in den USA

@itskarni @Broeki2 Kriege ich deinen League Account damit ich auch endlich mal höher bin als Gold?

Hallo ich bin gerade mitten über dem indischen Ozean, wenn ich jetzt abstürze vermache ich alle meine Anteile der NNO GmbH an @Broeki2 (Das ist was Schlechtes, keiner will die haben, und das wäre basically mein letzter Mittelfinger an ihn)

I’m on the $100/mo Claude plan. Tested the new “fast mode” with Opus 4.6: one task took ~11 minutes and cost additional $26. That cost/performance ratio feels way off.

@theo Have a good rest

No stream today, sorry. Tough day. Likely live on Monday to catch up on everything I missed.

@theo Sheeesh 🤙🏼

Posting so I can write off my broken board as a business expense

Just waiting for @riotgames to rework the kernel level anti cheat so I can finally move off of windows completely to play @LeagueOfLegends on @CachyOS peacefully 🫩

@convex Nice, congrats! But since you are a US company, you still fall under the US cloud act, don’t you? Just curious 👀

ok we finally did it 🇪🇺

Both did not complete the task fully. The goal was to implement logic and functionality into an app - not just the UI. Since this was somehow a frontend task, I will run both again on a heavier backend task to see, which one is ahead there. PS: Codex needed about 2x the time

I tested Claude Opus 4.6 and Codex 5.3 with the exact same prompt on the exact same codebase. Opus won imo on the design aspect again completely and it followed my instructions regarding the UI. Codex did not used the components I instructed it to use at all.