Vasu (@vasusrini) header
Vasu (@vasusrini) profile picture

Founder, 6hats,@skimotv Deeply interested in creating automatons that can construct models from anything in any context by directing attention, just like Humans

Silicon Valley
Joined April 2007
Photo by Vasu (@vasusrini)Photo by Vasu (@vasusrini)Photo by Vasu (@vasusrini)Photo by Vasu (@vasusrini)Photo by Vasu (@vasusrini)Photo by Vasu (@vasusrini)Photo by Vasu (@vasusrini)Photo by Vasu (@vasusrini)Photo by Vasu (@vasusrini)Photo by Vasu (@vasusrini)
Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5: hackertube.org/ - Excellent recs without any logins (state is kept in local browser session) - Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads - Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version - HackerTube Longs ✊ A bit more on the process: I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing. Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.
2,414
4,868
1,831
52,691
5,963,823
How do you teach a humanoid robot to walk without programming every move? Reinforcement learning step by step 1. Build a simulation (MuJoCo or NVIDIA Isaac Lab) with joints motors sensors and a physics model. 2. Define the task: walk forward while keeping balance. 3. Write a reward function points for moving toward the target staying upright and matching the desired speed. Penalties for falls, wasted energy and unstable movements. 4. A neural network, the policy, reads the robot's state (joint positions, joint velocities, body orientation) and outputs target positions or torques for each joint. 5. An algorithm like PPO updates the policy over many episodes and millions of simulated time steps. 6. Close the sim-to-real gap with domain randomization, then test carefully in the real world.
You're not ready for this. One of the most human-like humanoid robots has just entered production. Tesla says Optimus production has started on its high volume manufacturing lines in Fremont with the first production quality units coming off the line. Tesla says the current process is suitable for a commercial product, although the hardware and manufacturing methods will continue to evolve. The company has also outlined plans for a Fremont line capable of producing up to 1 million robots per year. That is a target capacity, not its current output. How quickly can Tesla move from early production to manufacturing humanoid robots at scale? And what production numbers will we see over the next 12 months? Source @ SawyerMerritt
4
3
30
15,629
A caterpillar that takes the form of a snake when it feels threatened. This baffles me because it’s not like this caterpillar consciously evolved this way on purpose. Evolution is crazy.
248
1,659
136
13,362
787,208
Congratulations to Samhita Vasu (’24 Biomedical Engr)! She won a Fulbright research grant to India to develop an at-home creatinine sensor to increase access to early screening for chronic kidney disease, working with mentors at Mehta's Children's Hospital, Chennai. #fulbright
1
2
1
7
636
Men are hunters and women gatherers, right? Not quite. Mounting evidence shows women are physiologically better suited for endurance hunting than men, thanks to the power of estrogen. Ancient female hunters could pursue prey for miles. scientificamerican.com/artic…
9
22
5
119
26,939
# automating software engineering In my mind, automating software engineering will look similar to automating driving. E.g. in self-driving the progression of increasing autonomy and higher abstraction looks something like: 1. first the human performs all driving actions manually 2. then the AI helps keep the lane 3. then it slows for the car ahead 4. then it also does lane changes and takes forks 5. then it also stops at signs/lights and takes turns 6. eventually you take a feature complete solution and grind on the quality until you achieve full self-driving. There is a progression of the AI doing more and the human doing less, but still providing oversight. In Software engineering, the progression is shaping up similar: 1. first the human writes the code manually 2. then GitHub Copilot autocompletes a few lines 3. then ChatGPT writes chunks of code 4. then you move to larger and larger code diffs (e.g. Cursor copilot++ style, nice demo here youtube.com/watch?v=Smklr44N…) 5.... Devin is an impressive demo of what perhaps follows next: coordinating a number of tools that a developer needs to string together to write code: a Terminal, a Browser, a Code editor, etc., and human oversight that moves to increasingly higher level of abstraction. There is a lot of work not just on the AI part but also the UI/UX part. How does a human provide oversight? What are they looking at? How do they nudge the AI down a different path? How do they debug what went wrong? It is very likely that we will have to change up the code editor, substantially. In any case, software engineering is on track to change substantially. And it will look a lot more like supervising the automation, while pitching in high-level commands, ideas or progression strategies, in English. Good luck to the team!
Today we're excited to introduce Devin, the first AI software engineer. Devin is the new state-of-the-art on the SWE-Bench coding benchmark, has successfully passed practical engineering interviews from leading AI companies, and has even completed real jobs on Upwork. Devin is an autonomous agent that solves engineering tasks through the use of its own shell, code editor, and web browser. When evaluated on the SWE-Bench benchmark, which asks an AI to resolve GitHub issues found in real-world open-source projects, Devin correctly resolves 13.86% of the issues unassisted, far exceeding the previous state-of-the-art model performance of 1.96% unassisted and 4.80% assisted. Check out what Devin can do in the thread below.
327
1,754
286
10,670
2,083,665
Replying to @davidchalmers42
I sure don't.
25
18
1
418
37,769
“Never delegate understanding.” —Ray & Charles Eames youtube.com/watch?v=q8vIESei…
1
11
66
10,135
A new band of researchers is finding that you don’t need 10,000 hours in a monastery before meditation can upend your entire psychology — and yet, the current body of meditation research has had surprisingly little to say about this middle ground between stress relief and enlightenment bit.ly/3qJPdk6
9
2
1
6
1,245
Don’t Be a Donkey TIM: “What advice would you give to your 30-year-old self?” DEREK: “Don’t be a donkey.” TIM: “And what does that mean?” DEREK: “Well, I meet a lot of 30-year-olds who are trying to pursue many different directions at once, but not making progress in any, right? They get frustrated that the world wants them to pick one thing, because they want to do them all: ‘Why do I have to choose? I don’t know what to choose!’ But the problem is, if you’re thinking short-term, then [you act as though] if you don’t do them all this week, they won’t happen. The solution is to think long-term. “To realize that you can do one of these things for a few years, and then do another one for a few years, and then another. “You’ve probably heard the fable, I think it’s ‘Buridan’s ass,’ about a donkey who is standing halfway between a pile of hay and a bucket of water. He just keeps looking left to the hay, and right to the water, trying to decide. Hay or water, hay or water? He’s unable to decide, so he eventually falls over and dies of both hunger and thirst. A donkey can’t think of the future. If he did, he’d realize he could clearly go first to drink the water, then go eat the hay. “So, my advice to my 30-year-old self is, don’t be a donkey. You can do everything you want to do. You just need foresight and patience.” *** Listen to my interview with Derek Sivers (@sivers): tim.blog/2015/12/14/derek-si…
60
329
54
2,017
360,560
Software estimates are one of the oldest lies we tell ourselves. We all know they don't work, but pretend they mean something and later feel enraged when shit hits the fan. I focused a big part of my undergrad on software estimation. After graduating, I wrote plenty about the topic. Then, I started working for a company where I spent years researching how to make better estimates. We sold multiple millions of dollars of software using the tools I built. I read everything there's to read. I could recite Steve McConnel's "Software Estimation" book from top to bottom. Here is the most important lesson I learned: People can't estimate software. It doesn't matter who they are or how much experience they have. Estimating software reliably is science fiction. And the best part: They will ask you to estimate something. They will tell you they understand it's not exact. They will promise they won't hold you accountable. And then they will. They always do. There are two solutions for this. Let's start with my recommendations for those who don't have a choice: 1. Remove "quick," "simple," "straightforward," "easy," and every similar word from your dictionary. Never use them. Don't let others use them when referring to your work. 2. Never volunteer an estimate. Everything you say will be used against you. 3. When forced, estimate work you know you can complete today. Always estimate with a range: "It will take me 2 - 4 hours." 4. Estimate anything you won't do today in days and weeks. Say, "I should finish that feature sometime this week." Do not estimate future work in hours. But we all know your manager will force you to give an estimate. Here is what you should do: 1. Estimate how long you think it will take you to complete the task. 2. Multiply the number by 3. This will be the lower range of your estimate. 3. Double the lower range of the estimate. This will be the upper range. Example: If you think something will take you 1 day of work, say "between 3 and 6 days." Here is the funny part: It won't take you between 3 - 6 days. This is as much bullshit as any other method you can think of. The true solution for this problem: Work for a company that doesn't care about estimates.
501
1,610
234
10,010
1,948,511
On the latest Huberman Lab episode, @tonyhawk and I discuss success in skateboarding (and life) and that it isn't always about being a 'natural.' It's about relentless drive, discipline, and the will to rise whenever you fall. Natural talent counts, but showing up, striving, and pushing through the friction — that's what makes the real difference.
33
97
13
1,163
258,078
Here's the full access link to my paper on the genetic code: rdcu.be/df4pl
3
22
65
11,638
If you want a problem solved, don’t form a team. Find the brightest person and let them work on it. Placing them in a team will, on average, reduce their productivity. Never form a team if there is one person who can sort out the problem. unz.com/jthompson/group-iq-d…
38
196
35
1,537
200,188
In 1974, author Ray Bradbury was asked, “What is space travel going to do for man?” In response, he gave the most mystical, mind-blowing, and strangely moving answer I could have imagined.
320
4,989
733
21,149
2,946,822
Rick Rubin became a world-class music producer without knowing the first thing about music
110
597
733
4,773
1,906,302
I asked 15 questions to #ChatGPT on weight loss and here is the "weight loss" booklet drive.google.com/file/d/1APd… #ChatGPTCuration
1
1
Forest tribe meets a white man for the first time in 1993
784
5,966
681
55,346
Sharing meeting links is so 2022. Your visitors simply walk-in to video visit you or call from Zoom or Teams meeting. Get early access by signing up now at hilo.video (pronounced "hee·low")
1
1