Helping the smartest people thrive past the Age of AI. LinkedIn and Stanford University. Humans will never be abstracted.

Ray Villalobos ✝️ retweeted
Introducing Gemini 3.8 Flash, another jump in Gemini's agentic + coding capabilities, and our 3rd updated Flash model in only 6 weeks... This model has been a ton of fun to work with, excited to see what you all think!
790
838
11,111
1,798,617
The way you get nerds to take a break is to take away their tokens. It's the new...The Internet is Out. I have to wait a whole hour now before I can do anything meaningful.
1
206
In my own head...I'm hilarious
1
195
Believe in yourself. No, I'm not talking to you. Stop doing that, this isn't that kind of post. I mean tell your agent to believe in itself. That's because Chatbots don't have a quality problem. They have a confidence problem. I've been disappointed at the output I get from coding agents lately. It's like my teenagers when I give them a task. It's fine, but I get the feeling their heart wasn't in it and the output is sort of mid sometimes. I'd read that Anthropic folks were literally sending Claude "believe in yourself" and "keep going" while it took a run at the Riemann Hypothesis. I know, at first I felt just like you're feeling right now. No way, that's ridiculous, it's not going to work. But heck I said, couldn't be worst than it is now. What the heck and gave it a try. (You can see the before and after) I've been building VibeGlossary so my Stanford vibe coding students can learn UI language. I wanted to add motion principles. Duration, easing, etc. The first pass was three dots sliding on a line. It wasn't even accurate, somehow it didn't even work. I had a few minutes before I went to church this morning, so I figured what the heck. Here's what I told my bot. "The motion teaching sucks. It's too small and not interesting. Be way more creative. You can do it. I believe in you." I felt funny about it. It sounds ridiculous. What did I have to lose. When I came back I had the second screen. I was floored. Unbelievable difference. Maybe I need to dust off Tony Robbins. Srsly, I think he'd make a killing motivating robots. It's not the first time I've told my agent to work harder and it's worked. Which makes me wonder. Do they just have a confidence problem, or is it a laziness problem. Don't know, don't care, better output FTW!
224
Everyone, stop trying to make Omarchy a thing
227
Now when I watch a video and its from like 4 months ago it reads like ancient history
1
225
Perplexity is the tool I wish I could use, but I've got so much invested everywhere else, and they all have me with their sweet subsidized tokens. Dang you Claude...you figured me out.
220
Kudos to the people naming the pre-release models. Ox Alpha sounds so based. I mean who wants feline beta.
205
A new harness strategy arises with Grok Bot. Cody (developer) makes changes, Tess (tester) verifies things are aligned to agents.md rules, Desi (Designer) reviews alignment to design.md. Every day I'm running a new Loop where Tess checks one of my sites and automates the process.
2
232
This is going to be awesome or suck...We'll see. It already filed 10 issues.
1
49
The worst thing about using OpenClaw with Telegram was getting a ton of random spam messages. I cant wait to get off it.
1
230
The new memory in Claude is a whole new ballgame. The syncing between web and Apps is sublime. That was one of the disappointing things before. Now, it feels like a whole new tool
211
Just when local models got less interesting, Apple calls the Mac mini an always-on agent box. Man, I think they missed the boat. I already lived that. Otis ran on a Mini I left on at all times. Grok Bot replaced the always-on box, just as RAM went up almost 5x in a year, so stuffing a desk with local memory is just not worth it. I love my mac mini. Otis will handle something else in the future, but I can't afford the memory pricing right now, so it's a no go for me. Way to miss the moment
210
I've gotten to the point that I've taken it for granted that I'm talking to an AI that can understand and do things. I'm not trying to get it to answer simple questions. I tell it to do things and it does them. The weirdest thing is that no everyone's doing this.
200
The students in my Stanford Vibe Coding class are really rocking some sweet builds with @GeminiApp AI studio.
165
The new @claudeai memory is legit awesome. My favorite thing most people aren't talking about is that they folded whatever you did in the chrome extension so that it now has access to the rest of your skills, chats...everything. It's what more useful.
213
Ray Villalobos ✝️ retweeted
We didn't forget about @Bot either... It's now available for all paid customers. $20/mo to change how you work with computers! We've reset weekly usage limits for all Bot customers. Try it out!
We just increased the included usage of Grok models in Cursor! Demand has been increasing with the launch of Grok 4.6. We already doubled limits last month with more compute. And now we're raising limits again permanently. Enjoy!
141
58
1,722
142,674
Ray Villalobos ✝️ retweeted
Gemini 3.5 Transcribe has arrived! Features real-time streaming, multi-speaker support, and Smart Mode to easily convert raw audio -> polished text Used it to build "Mind the Task" for my commutes 🚅 I speak whatever is top of mind and it gives me a clean, actionable to-do list even over the chaos of NYC subway noise
3
4
74
4,009
You're not the Empire. Don't fire the expensive laser at every file. Send a scout. Keep the hero for the trench. Token-min, not tokenmax.
1
178
I'm wiring that into a CrewAI crew this week. Cheap models on the volume. Frontier when it counts. A verifier that actually runs the tests. Don't put the flagship on lint.
1
24