Russell Smith 🤙

Having done YC before, it's amazing to be back. The energy is the same 14 years (!!) on. Also, the quality of the folks I met was just as a high, if not more. So many have done a ton already, some are early. All are moving super fast. Crazy achievement by the YC team and @garrytan. My personal takeaways from the kick off of P26: - the vibe/how folks/users feel matters - prolly even more for hardware - what is the smallest thing one can actually ship rn - watch out for hero-mode; are we solving the real problem, or the easy one - strong cofounder bond is key, and the easiest way to fail, even vs funding - building a pro sports team, not family - post more to LinkedIn, as spaced rep is a thing, and starts a brand - create a tribe. convert people. if founders don't no one will: but build one tribe, not more

Tech like this, fiber and theseus are coming. How we stop them is behind.

@CostSegDylan Isn't hearing also excluded lol?

Why the hell is dental or vision not included in healthcare? Like…the ENTIRE rest of the body is covered…but these two random things…nah. These are expendable

@xoaanya Ask PMs, CEOs and designers everywhere; can they really call those projects they worked on theirs?

Can you really call a vibe-coded project yours if you didn’t write a single line of code?

youtube.com/watch?v=EtK5es… by @Ken_Goldberg @MITRobotics 100% agree that the future of robotics lies in combing deep learning and foundational engineering principles. good old fashioned engineering + data = win.

@ronhung @paulg Love this way of putting it. We did venture debt at a previous co; paid it off ok, not sure it was worth it in retrospect, but didn't kill the co.

@paulg Seems obvious but don't give someone else a deadline for your company. Come to think of it, make a PG agent in your Claude Code that'll whack you on the head when you mention it.

Avoid venture debt.

@garrytan Also means I can leave autoresearch / other things running whilst ooo/afk.

@garrytan I use tmux + mosh + tailscale to another machine, more beef, 10g net, h100. I use my laptop as an interface, or phone to check in on things. Laptop stays cold.

Ridiculous problem: I am looking at my MBPro (often running CI tests and evals locally) all the time in cars, planes, and just waiting someplace I need an attachment to the bottom of the MBPro that deflects heat so my lap is not burned all the time. I don't want to carry a big lap desk, that's very awkward. Anyone know of a good solution? Anyone make a 3D printed option for this?

@garrytan Does gstack have a new skill for it? 🤞

Garry’s List is now a 95 on page speed and the codegen luddites should take a seat 🫡

@a16z nobody gatekept shit x.com/michael_kove/s…

@Crislycai @a16z Think this is a great point. It's judgement, and taste.

The constraint now has shifted from engineering to judgment. When the technical barrier collapses, the real differentiator becomes the clarity of the idea, the criteria behind it and the ability to direct a tool instead of depending on a specialist. AI democratizes execution, but it also exposes who actually knows what they’re trying to build.

20 million developers used to be the gatekeepers to software. Wabi CEO Eugenia Kuyda on who gets to build now:  "The only people who could make software up until last year or so [were] just professional developers... Very few people in the world." "Even if you have a good idea, go ahead, find an engineer, find a co-founder, find a technical co-founder to build this. But you can't really otherwise build it... It requires a lot." "But now anyone can." "This is just wild." "Before  we had to spend months developing that, developing an app and figuring out some illustrations and sounds. Now you just write one prompt and it's there." @wabi CEO @ekuyda on @solofounders with @julianweisser

"But now anyone can." Kinda. It's still currently high leverage if you actually know what you're doing. I.e. you will build slop easily otherwise.

A lot of guesses, but passive is believable. We develop passive sensors for this very reason - emission = possible to track, therefore easier to find.

This is sick. I've been doing a less pro version of this with Obsidian and OpenClaw. It's awesome for cron'd tasks, and getting briefs daily, plus isolating it from my main stuff. I send context I want via iCloud Drive read-only, and receive back via the same from its account. Excited to learn from yours!

LLM Knowledge Bases Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So: Data ingest: I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them. IDE: I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides). Q&A: Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale. Output: Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base. Linting: I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into. Extra tools: I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries. Further explorations: As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows. TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

If you love your work you don’t work for a day in your life

@Quiettangent @MarioNawfal think that's why he left.

@MarioNawfal What is this Myron dude is talking about?? This narrative does NOT apply to Iran!! Omg stop spreading propaganda. You can be critical of the US but do not spread IRGC propaganda!

This isn't good. Also hearing about bounties. Hope the search and rescue finds them first.

BREAKING - DEVELOPING: American fighter jet shot down, and Iran is trying to capture the pilot while the U.S. launches a massive search and rescue operation Seems Iran shot down an F-15 fighter jet, and is trying to captured the pilot A massive American search operation is underway involving MQ-9 drones, F-35s and heliborne units Iranian air defenses are engaging the American aircraft involved in the search and rescue Location geolocated to the Khuzestan Province, southwest Iran – 30.6515, 50.1174 This would be the first time a country intentionally shot down an American fighter jet since 1999, 27 years ago, and the first time since the Iraq war a pilot was captured (in Iraq it was 2 Apache helicopter pilots) Iranian state TV just put out t hat statement: "Anyone who captures the US pilot alive will receive a valuable reward" This remains unconfirmed, but many indicators are pointing to this being true CONTEXT: The F-15E is worth about $80-100M, and they have NEVER been shot down. They are a very difficult aircraft to shoot down, and usually requires luck + surprise. This demonstrates Iran's military capabilities are far from destroyed, and this successful downing of an F-15 may have involved the Russian S-300

Our military installations have been turned into gun-free zones—leaving our service members vulnerable and exposed. That ends today.

@karrisaarinen @jordihays @tbpn 2) what?! Insane. Congrats @johncoogan!

@jordihays @tbpn 1) what

TBPN has been acquired by OpenAI The world is changing quickly but TBPN will stay the same. Live every weekday just with a lot more resources. Thank you to everyone that has been a part of this journey big or small. We are 17 months in and unironically just getting started.

Still nothing stopping these.