Alex Key

من يدري إن كان الجود لا يتلاشى الصفاء مع كل فجر؟. AlexKey12🎆

@specs the artistry in this is honestlyy next level

See your world through a new Lens with SPECS, augmented reality glasses that make computing more human. Pre-order now.

alright so hear me out, i waved back at someone who was waving at the person behind me not sure what to do with this information

rooftop views on the outside, aura

@0xCarnagee the vibes are simply immaculate today

OpenAI engineer: "the cost of putting up a plausible PR has gone to zero. the cost to review it has stayed the same. close the loop. give the agent a way to verify its own output." in an 82-minute podcast, the founder of ruff and uv explains how software engineering actually changed. codex + automated verification + green CI - that's the new bar. Watch the talk

@NYSGTSC what is the name of the song playing in the background

Whether you're riding an e-bike for your first time or your 100th, learn how to have a safe ride every time.

comfort isn't about being expensive. it's about being deep

@0xrevayz i found the original uncut version of this and it's actuallyy 10x crazier

Anthropic engineers just built a full app using a loop of 3 agents. The team behind Claude Code proved the winners of 2026 won't have the smartest model. They will have the best loop But a loop on top of an empty workspace is just expensive noise Most people open a blank chat, paste context, get a generic answer, and start over tomorrow. That is not a model problem. That is a setup problem A real Claude setup runs on 7 layers. Preferences kill repeated instructions. Styles save operating modes. Projects keep context from mixing. Knowledge files load real material. Connectors plug into your calendar, tools, and repos. Artifacts turn answers into usable tools. Workflows make repeated tasks run the same way every time Same model. Different setup. Completely different result To most people Claude is a chatbot To architects Claude is an operating system Stop visiting Claude. Start living inside it

@rojda_havin you are unbelievably prettyy ❤️‍🔥...

my life👏👏👏

@therealfuguy i swear you get prettier every post

This day has been just splendor!! Petite freshman year hookup is transferring #AG2G

enough said

today was one of those days where everything just felt right for no reason

@ScottyBeamIO The way I just gasped out loud in public because of this

A GUY GAVE A HERMES AI AGENT ONE GOAL AND IT GENERATED $22,000 IN LEADS FOR A STRANGER'S BUSINESS IN 24 HOURS On stage at a conference, he picked a random small business from a LinkedIn post in the audience. A web design studio in Carlisle, Cumbria. He gave Hermes Agent – open source, runs on any LLM – one goal: "Find companies in the area. Take screenshots of their current websites. Redesign every single one. If they don't have a website, build one. Email each business owner offering the redesign for £2,000." The agent went off and did it autonomously. Hundreds of full website redesigns. Real ones. Not AI slop – actual professional-looking sites. It could only send 50 emails a day to avoid looking like spam. 100 emails sent. 11 responses. Roughly $22,000 in leads sitting in one inbox. From one prompt. For a business that didn't even know this was happening. The agent didn't need a team, a sales department, or weeks of outreach. It needed one goal and permission to keep going. Bookmark this post. Full demo in the video below.

some chapters need to close before better ones can begin @hospitalhospita 👀

@MyWestLord this hit me at exactly the right time and i needed it more than i realized

4 AMD nodes, 1 rack, 8U tall, all to run AI without ever touching the cloud. He stacked 4 Framework desktops into a mini rack half the width of the one beside it. Each node: its own fan, its own CPU cooler, its own power supply. A 5 gigabit switch ties them together. Ansible wires them into one brain. A Beowulf cluster, the kind universities ran in the 90s, now sitting in a spare room. Every watt feeds into Home Assistant. He watches the power draw live, model by model, asking the same question every time. How cheap can I make a machine that thinks? The rack hums. 4 chips split the load. Large language models run on hardware he owns, no subscription, no data leaving the building. Then he says the part nobody wants to hear. A single Mac Studio still beats the whole stack. One box. No cables. No Ansible. No 8U of metal. It runs the big models faster and cheaper than 4 nodes glued together. Months of wiring. A rack you can point at. A cluster that boots from a script. Beaten by something that fits under a monitor. He built a server room to outsmart the cloud. Apple sold the answer in a box.

feminine energy, boss mindset

boreddddddd

I hate studying for tests and plus im horny

@BadBitchKCHeinz thanks man