agentic coding pricing is unsustainable at a flat $/message structure. a power user in a session will burn about $10+/hour in tokens. so flat per-message pricing starts off egregious and then margins go negative as the context window fills up during convo bc of increasing input

Ethics aside, 50/50 is a prodigious split. It's even better than syndicated crime and 🌶️ work.

Imagine switching to a superior product and avoiding an egregious rug pull in the same move.

there's this phenomenon happening with AI coding tools. i'm calling it "subscription fog." you have users bragging about getting $3k of inference for $200/month. then suddenly things "feel" off. when the same entity controls both: 1. the inference supply 2. the harness

TIL that men have a ticking biological clock as well. And it's comparable in timelines to that of women.

Bring memes to life with the @Grok app!

This is incredibly funny to me, because, a guy at my previous workplace was fired for posting this meme in the company chat; he was protesting something the moronic admin was forcing down our throats. Field day for HR lol.

Despite the Gemini team training an excellent frontier model, I can't help but feel that Google is lacking direction on the AI product side. Why can't I give Gemini agentic access to my Gmail? Why can't it read my Google Docs to get the context it needs? Why can't I ask it to do

(Me): It's MoE model, two experts. One is aligned, and the other is misaligned. (VC investor): go on (Me): The kicker: it's a wearable. The aligned model sits on your right shoulder and has a halo. The misaligned model is on your left and has horns. (VC investor): oh my GOD

Not sure if it's meant to be a joke, but I have actually done this in prod, to great success. Back at Vuala, our agent boxes running browsers would often eat up all the RAM and OOM. Wasn't consistent, nor could we preduct RAM usage. Auto restarts with Cloudwatch were the way 💪

posted about this a while ago but even more convinced of it now i think LLMs are going to plateau right at the point where they make programmers lives even better than they already were but fail to eliminate them

This is such a wild & reactionary take on what's going on here from Garry Tan - for someone who is usually so well-spoken, rational, and technology-forward, I'm shocked. Determining the identity of an agent, service, machine, computer, or other entity is a fundamental problem

Just found out about Amp! How is this different from Cody AI? Sourcegraph 👀

Too little too late. Already jumped ship to Apple Music.

For everyone reacting to that McConaughey clip about LLMs: THIS is what he wants. Not only can we not get any coherent output from such a short corpus (with current arch), continuously updating it at the pace of human mind drift is a monstrous challenge.

Did you know that when they say stuff like "The A18 uses TSMC's 3nm process" or "announced the 2nm node" The 3nm, 2nm actually doesn't mean anything?! It's just like a version number. They make it up. Literally nothing measures 2nm or 3nm. I certainly didn't know.

when aligning icons with text, you can set the height to 1lh it equals the element's line-height, so icons align perfectly

Here is my patented universal post incident review, all you ever need

Guys it’s 2025. Elders still have back and knee problems (as they’ve had for centuries before). Stop focussing on unrealised performance and optimise for your loved ones’ quality of life instead.

Production went down. Load balancer health checks failing. All instances marked unhealthy. SSH'd into an instance: - App: running - CPU: 5% - Memory: 40% - Disk: 15% - Network: fine Manually hit health endpoint: curl localhost/health {"status": "ok"} Worked perfectly. Checked