Stephen Lacy

@WilsonKyi @bchesky SpaceX as emerging? Ha

@bchesky "mafia recruiting" works for hiring the best people. It's the same method I use for investing in the best founders. By the time you open a role, you're already too late. The best people are always tracked. It's just a matter of who's been tracking them. Most founders run reactive searches. Role opens. Hire a search firm. 50 profiles, 10 calls, 3 finalists. Pick the least bad. A year later you realize they're mediocre. Brian does the opposite. He tracks talent mafias. In the past, many have been proven to be high talent density, and each of them have something their known for e.g. technical, ops, design, FDE from @Square @Uber @Dropbox @stripe @tryramp @scale_AI @PalantirTech @Airbnb Two ways to build a shadow list of the best people you're tracking (similar to @cursor_ai's hiring method) 1) Start with results, work backward to people. Don't say "Nike has good marketing, let me poach from Nike." Find a specific ad you love. Track down who actually made it. 2) Pipeline before you need it. "Who are the 2-3 best people you know?" The best VC's do the same thing. We don't wait until a founder is raising, because by that time the round is effectively done. Here are some of the AI mafias I'm tracking, and DM me if you want to know the rest of the list and how I prioritize based on specific teams, vintages, and function. @OpenAI @AnthropicAI @harvey @mercor_ai @cursor_ai @xai @SpaceX @SierraPlatform If you're a founder and you can't name the next 5 hires you'd make, or the next 5 founders you'd invest, someone else is already hunting them down. h/t recent AI founder mode episode on @InvestLikeBest @patrick_oshag

@Lazi_Learner @__Rhodium__ It could fly but we were indoors with limited space so we didn’t want to let it accidentally kill someone lol

@__Rhodium__ Was it too heavy to fly on it's own?

For the defense tech hackathon, I built a sabotage drone with a robotic arm that takes natural language commands and executes missions autonomously. -On-device LLM for reasoning, mission planning, flight control primitives, and success verification -Closed-loop visual servoing for stable positioning -Precise arm manipulation from a monocular CV pipeline: vertical kernel + PCA for axis and depth estimation (for antenna bending)

Was epic building with @__Rhodium__ on this crazy project!

@nicole_clash That could work for capturing demo and project details: “does the repo Actually Do Anything?” The results would be filtered down, not by taste/judgement but by quality. I would also have the audience vote grand prize, that is how I ran my hackathons.

Yo I was just thinking, isn’t the solution to slop judge inconsistencies at hackathons literally just having an aligned AI be a judge that sits in on all demos alongside human judges? Say you have a first round to figure out who the finalists are from out of 300 people, 6 people are finalists, and you have 60 judges. Instead of splitting the judges into groups of 3, and having each group do 15 teams of 3 minutes each, which is downright retarded (3 minutes is not nearly enough lol), You have your 60 judges evaluate 5 demos each, individually, each up to 10 minutes (teams don’t have to take that long if they don’t want to, but more explanation the better no?). You have an AI that evaluates all 300, consistently. Top 2 scores from the AI side makes it into finals automatically, the remaining 4 spots can be any system of weighing human vs AI opinions that you think is fair. Not proposing these exact numbers and weights, but you get the idea.

Ah actually, nvm. The judges were fking retarded. Just found out the one guy I talked to that had a sick project I was 100% fine losing to (I literally coded for 28 hours straight to get this shit built, didn’t socialize at all), they got 5th (I left after not getting past the first round). Took a closer look at the teams that won. What the FUCK lol. You’re telling me, my project, Where I literally had a fully working system, That ingested 10+ different data sources, Resulted in 1000+ military grade reports, With REAL LIVE data, MESSY data sources too, like literally Twitter, Didn’t get at least top 2, where every other project besides the one guy I had talked to, Is significantly worse in terms of real world applications and use cases? And would have taken me 2 hours to build? You’re shitting me, right?

@nicole_clash Yeah my teammate and I built a flying robot on a drone with a working autonomous arm, didn’t make it to the finals. I am salty

@shaggysurvives It’s pronounced sly

going to start pronouncing CLI like “clih” the way people pronounce GUI like “gooey”

@stevenzhang I did it on a: car, motorcycle, bicycle... haha

I did your drive

The deobfuscated code:

It is extremely annoying how easy it is to spread malware. Assume all Polymarket or crypto "trading bots" on github are malware. Example: It only takes 5s to realize the code smells bad. Let me walk you through it: 1. The readme says "No wallet or API keys required" < that is impossible - all trading bots must use either a private key or an api key. 2. This is a suspect import: CashBlazorLab/polymarket-trading-bot src/index.ts#L10 All other imports are `import` while this is a createRequire. 3. The `bn-eslint.js` makes no sense - no package should EVER import an eslint package in runtime. 4. bn-eslint.js is a fake package that hides it's true intentions: a. it reports as "big.js" but is called bn-eslint.js b. it imports the payload package `npm-eslint-helper` which contains the actual malware c. socket reports it as malware: socket.dev/npm/package/bn… 5. I downloaded the tarballs for bn-eslint.js and big.js 6. A quick glance of bn-eslint.js showed a weird import at line 605 where it imports `npm-eslint-helper` 7. I downloaded the tarbal for `npm-eslint-helper` and whatcha know, it's obfuscated code. 8. It grabs the .env, id.json, config.toml, and sends it to eslint-helper.vercel .app

@DefankoB @Atenov_D @Polymarket @PolymarketTrade yeah, its malware: x.com/stephenlacy/st…

@Atenov_D @Polymarket @PolymarketTrade typosquatted dependencies that are not legitimate npm packages: "bn-eslint.js": "^8.0.5", ← real package is "bn.js" "web-http-errors": "^4.5.1" ← real package is "http-errors"

This Polymarket arb bot made $317 profit without holding a single position for more than 4 minutes. No wallet. No API keys. Just live read-only prices and a 46-second entry window. > It targets 5-minute BTC Up/Down markets. Entry only when price lands between 0.14 and 0.26, between second 25 and 71. Miss that window - the bot logs "Entry window passed" and waits for the next one. Exit logic is the real edge. Minimum 3s hold. Defers sell if bid drops below 85% of entry. Forces out by t=265s regardless. 1% fee + 0.5% slippage already baked in. $100 in - ~$40 profit $500 in - ~$300 profit $1,000 in - ~$500 profit TypeScript. Open source. Logs everything to console and logs.txt. github.com/CashBlazorLab/…

@AlterEgo_eth @Atenov_D @Polymarket @PolymarketTrade It is malware: x.com/stephenlacy/st…

@Atenov_D @Polymarket @PolymarketTrade It's strange that they made it publicly available

@Atenov_D @Polymarket @PolymarketTrade 7. Obfuscated code:

Do NOT run or clone the repository - It is malware! It only takes 5s to realize the code smells bad. Let me walk you through it: 1. The readme says "No wallet or API keys required" < that is impossible - all trading bots must use either a private key or an api key. 2. This is a suspect import: #L10" target="_blank" rel="nofollow noopener">github.com/CashBlazorLab/… All other imports are `import` while this is a createRequire. 3. The `bn-eslint.js` makes no sense - no package should EVER import an eslint package in runtime. 4. bn-eslint.js is MALWARE: a. it reports as "big.js" but is called bn-eslint.js b. it is literally malware: socket.dev/npm/package/bn… 5. I downloaded the tarballs for bn-eslint.js and big.js

@Atenov_D @Polymarket @PolymarketTrade 6. A quick glance showed a weird import at line 605: I downloaded the tarbal for `npm-eslint-helper` and whatcha know, it's obfuscated code.

Won best edge AI at the @ycombinator and @innate_bot hackathon! We built a local VLM multi-rover orchestrator for Mars exploration. On-device navigation and automated fault detection & recovery across odometry, stereo vision, and lidar. Thanks for hosting, @ax_pey!

@iamjasonlevin Banger as always

I'm renting out a movie theater in NYC to watch IG Reels. The date? April Fools. But this is 100% real. Come out and enjoy IG Reels on the big screen where they belong

@0xTengen_ @PolymarketSport @Polymarket Low transaction count? Ha check the activity tab

polymarket trader made another $500k on sports and his pnl is now sitting at $11.3m this is an absolute record on the platform the guy is averaging over $1.1m in pure profit every single month he actively buys "no" on favorites because it covers both the draw and the loss we saw exactly this with his villarreal and psg positions his read on the top 5 football leagues is flawless like i said before, this isn't some hft bot, his low transaction count indicates that there's an human behind the screen also he snipes coinflips at ~50¢ in us sports: catching nba spreads and nba totals (knicks/pacers under 227.5 at 50.3¢) a masterclass in diversification and heavy sizing you can check his active and closed predictions below his profile: @kch123?via=tengen" target="_blank" rel="nofollow noopener">polymarket.com/@kch123?via=te… in a sea of bots, the biggest whale is still human

@taydotfun That is if the L isn't broken

the real size of new york is "30 minutes from your apartment" williamsburg is hard to beat

@BenniKim GTC vs FOK orders have different fees

Having fun with the polymarket dataset. Good sentiment indicator and also offering arbitrage opportunities to the sharp quant ;) I never though I'd use simplex and geometry to rectify probabilities and seek arbitrages on stuff like 174 contract wide PGA Tour golf tournament winning odds. Because there HAS to be a winner. Seems the implied belief volatility concept holds water too.

@disarto_max banger

Deployment details