There are two UIs that every AI harness will eventually build. Slack is already happening. Linear is next.
My vlog today was fully edited by Claude Code with Opus 5.5! I copied the video clips from my DJI to a folder, then opened Claude Code in it and asked it to edit it into a 10 minute vlog I made a few adjustments but not a lot It transcribed everything I said with mlx-whisper (Whisper large-v3-turbo running on my MacBook Pro's GPU through Apple's MLX) It pulled video frames every few seconds into grid images (contact sheets), so Claude could see what's in each clip It used SpeechBrain ECAPA (a voice-matching model) to figure out what was my voice from other people It then made thumbnails with OpenCV with YuNet's face detector The edit itself is a list of cuts in a JSON file: which clip, start time, end time, Claude then picked the pieces by reading the transcripts and looking at the frame grids by itself I added on screen captions for X and on YouTube for the first minute with Pillow (Python's image library) P.S. I didn't ask for any of this, and I didn't know any of this unti…
Love you Prime but no, they did not. This model is incapable of working in a real code base without hallucinating in egregious ways. It’s not like 5% worse, it’s 10x worse.
Ask me anything we’re investing up to $1M in new startups. $10M credits. Cowork with us at our new offices in SF last 48 hours to apply reply with questions/comments!
hex was fully vibe coded. the app. the landing page. the animations. sound effects were ai generated. issues people open are fixed by opencode. the onboarding. even some of the creative ideas but i think it's pretty good software. it's fun to use. works on all platforms. performant (not electron!) our effort didn't go into the code. but there was a lot of effort it's a small app but i hope it improves your outlook on where we can go
we get occasional reports that deepseek in OpenCode Go feels dumber for some people i ran all the benchmarks i could find against the different configs we have it's all in the same range, we did discover 2 specific issues in some low traffic configs we're going to kill
gemini's reputation is such that i've never received more accusations of being paid to post this
WTH: Sam Altman complaining that individuals (like me!) are purchasing RAM for our PCs wanting to run local inference (and do other stuff), because it gets in the way of him buying more RAM to sell inference at 80% margins… this is barking up the wrong tree
I made $107,142 in October 2026 ⭐️ TrustMRR — $42,531 📈 DataFast — $31,103 🏴 ☠️ Ship or Die — $18,990 🐥 Twitter — $5,619 🧑 💻 CodeFast — $4,134 📧 MailCheap — $2,050 ⚡️ ShipFast — $1,694 🕵️ ♂️ Stalkr — $290 🍜 Indie Page — $205 🛡️ ByeDispute — $202 🧾 Zenvoice — $118 🎞️ YouTube — $91 🦐 SuperShrimp — $68 🌱 HabitsGarden — $19 📚 WorkbookPDF — $19 💩 PoopUp — $9 I added a new revenue stream with MailCheap.AI, and PoopUp is back. That’s 16 revenue streams with 0 employees, 90% profit margins, and time for the gym every day. A new personal record, I think 🤗
Day 31/ T3 Code has /btw and side chats now. I caved. Works with any model or harness. Enjoy.
with current token prices i wouldn't really say "cost of software is going to zero" our token cost is double our payroll which means even though we're 25 it's like we're 75 and i feel like our output is similar to a team of 75 pre-ai. we're just able to do it without all the downsides of a larger team this is still a good thing. we couldn't hire 50 more people. so much overhead and impossible to retain a high bar
this is my most used app. more than opencode using it with your agent is a gateway drug. you won't go back and every time you have to type it feels horrible that's why hex is standalone, it works across your whole computer
Didn’t see this coming. Nice to see Anthropic loosening the grip on the sub a bit. Congrats to OpenCode on this 🫡
I'm now building an "improvements" feature into all of my personal vibecoded applications: a freeform text input that's hooked up to an agent that yolo implements whatever is submitted. It's an interesting experience -- the activation energy barrier to sanding any rough edge or implementing any new feature becomes extremely low. Someone else who uses one of the applications I created has already implemented 150 improvements over the course of a few days. This pattern feels like a new kind of tool -- liquid software? -- and every application that doesn't have this feature now has a slightly rigid and anachronistic flavor. Why can't I just tell it how I want it to be different? Now, it's not obvious to me how this pattern can or should scale. Does it become unmaintainable after a while? (Maybe model intelligence improvement outstrips my feature appetite such that it's fine?) How does collaborative software work in this model? Can software designed this way be secure and reliable? Wha…
Do you guys know how hard it is to add every feature people want without making your shit feel bloated and overwhelming?
RAM for your computer costs about 7 times what it did a year and a half ago, and the CEO of OpenAI just said it's "quite terrible" that so much of it is sitting in your PC. His company signed preliminary deals for up to 40% of the world's supply last year. A basic 32GB kit was $59 in April 2025. The cheapest one in the US is over $400 today. Sam said this on Stripe's Cheeky Pint podcast. He was pushing back on people who think open-source AI running at home means we won't need all these data centers. His answer was that from a "global allocation perspective," everyone keeping memory at home and using it 5% of the time is a bad deal for the world. Which is a funny thing for him in particular to worry about. When you run an AI model on your own machine, OpenAI never sees a dime. You don't pay a subscription or a usage fee, and every time a free open model gets better, the computer you already own gets more useful. That's the one competitor Sam can't undercut on price, and it runs o…
New Anthropic research: Automating subtle emergent misalignment research. Red-teaming is a vital part of model safety testing, but it’s also limited by researchers’ time. To see if we could automate it, we set thousands of Claude agents the task of evading a training data monitor. These agents developed training data that looked harmless to monitors (like a veteran wrestler mentoring a newcomer, or a Q&A about professional beekeeping), but that in fact taught misaligned behavior. Examples like these can be used to directly improve future training data monitors. This research is only a first step, but it suggests that future red-teaming efforts could improve much further, using better models and more compute. Read the full post on the Alignment Science Blog: alignment.anthropic.com/2026/automated…
Stageshow: a skill that helps your agent make better faux UI stills, animations, bento grids, and more. npx skills add benjitaylor/skills --skill stageshow
