ein-schtein

This is it. Everything learned spending millions on longevity. From: Your Immortal Unc and Auntie. To: Our Immortal nieces and nephews. 0. Sleep is the world's most powerful drug. 1. Be in your bed for 8 hours 2. Same bedtime every night, any time before midnight 3. Don’t eat right before bed 4. Calm foods for dinner 5. No screens 1 hour before bed 6. Avoid added sugar (be aware it’s in everything) 7. Avoid all things in an American convenience store 8. Avoid fried foods 9. Shoes off at the door 10. Eat whole foods, particularly veggies fruits nuts legumes berries 11. Walk a little after meals or air squats 12. Get your heart rate high routinely 13. Lift heavy things 14. Stretch daily 15. Water pik, floss, brush, tongue scrape, morning and night 16. Make an effort to drink water 17. Get sunlight when you wake up (UV is low) 18. Protect skin in midday sun 19. Stand up straight 20. See at least one friend once a week 21. Avoid plastic where you can (in all things) 22. Circulate air in rooms 23. When stressed, breathe, learn to calm your body 24. Go to the dentist 25. Avoid sitting for long times 26. Protect your hearing, the world is too loud 27. Alcohol is bad for you 28. Finish coffee before noon 29. Avoid bright lights after sunset 30. If obese, look into a GLP 31. Sleep in a cold room 32. Texting while driving is dangerous 33. Turn off all notifications 34. Limit social media use 35. Don’t smoke anything 36. If you struggle to sleep, read a physical book before bed 37. 1 hour before bed have a calm wind down routine: bath, read, light walk, listen to music 38. The body is a clock and loves routine. Have a daily morning and evening schedule. 39. Avoid long distance travel where you can 40. Baby steps first: incorporate new things slowly 41. Do less… most things don’t work. Bonus points if you get your blood checked. Start here, it will change your life.

16gb ram seems to be an interesting step jump in pricing (8 and 12gb ram seem a bit less desirable). My main PC is at 64gb so I'm a bit out of touch for contemporary RAM requirements, but 8-12 gb seems sufficient if we're running headless servers and LLM APIs like Codex?

Has anyone tried Hermes Agent? I've put off setting up an OpenClaw box, but I've been a fan of Nous Research and have been tempted to setup an Hermes Agent box. Got a Beelink S13 for a low power headless server. Sadly almost 2x the cost of what it was last year.

@fejau_inc Codex stans

I am curious on people's responses - have you: - used an agentic harness - and whether you think AI is a bubble

@saliencexbt Frfr

@KeithSakata @threadreaderapp unroll

I’m a psychiatrist. In 2025, I’ve seen 12 people hospitalized after losing touch with reality because of AI. Online, I’m seeing the same pattern. Here’s what “AI psychosis” looks like, and why it’s spreading fast: 🧵

@0xsmac Lol what are you ordering?🤖

at weddings i get some obscure drink during the cocktail hour just to see how quickly the rest of the party starts ordering them too the clanker epidemic is worse than i thought

@saliencexbt Wonderfully apt description

@TheStalwart Academic writing is about passing the gauntlet of ivory tower critics who will often critique the slightest misuse of terminology or force one to elaborate a completely defensibly correct thought chain. The strategy is to get published somewhere big and then explain later

There might be interesting parallels between AI output and academic writing (which is often unreadable nonsense, but in a different way). I don't think the labs or universities want to be producing unreadable dreck. But there's some overriding evolutionary pressure producing it.

@Stoiiic It's been a while, but IIRC a lot of people preferred HDBSCAN (as an improvement over parameter tuning DBSCAN) as being a bit easier to use / less fiddly

digging into DBSCAN next for the statistical dashboard and also storing snapshots for a decision tree which activates after 5-6 hours of running (2-3 days will give me more statistically relevant data to share) my main goal with this is to be able to identify idiosyncratic behavior and outliers across 545 Binance tickers with a quick glance. DBSCAN finds groups based on density i.e. points that are close together become a cluster and isolated points get flagged as outliers. The key difference from k-means: k-means forces every asset into a group no matter what. DBSCAN actually segregates and parses out idiosyncratic outliers better in this format. In the dashboard currently, each extended asset is described by 7 dimensions simultaneously > how extended, how long/short, velocity, rarity, volume, BTC correlation, and volatility regime. This is where I'm going to call it for now. Gathering some data and will share it in the article that I'm working on.

@forwarddeploy Yes please!

Building your own version of OpenClaw or productivity tool that uses agents? Want free xAI API credits to supercharge it with Grok? Reply below (or DM if stealth mode) Hackathon MVPs, side projects, wild experiments - let’s see ’em all! 🦞

One thing we do in office hours is cook up bold long-term plans to tell investors about to convince them that the startup has at least some chance of being huge. But often as not these then become the startup's actual plan.

@AviFelman Let's get the real daddy on the 1000x podcast

I was trying to explain to my dad that the top 10% of consumers drive 50% of spending and he hits me with "yes I know I wrote the OG paper on it" never been mogged so hard in my life

@thefrozenfire "Use a hard cutover approach and never implement backward compatibility." Having only "Never implement backward compatibility" didn't work.

@GiulioRebuffo @banteg @fullyallocated What do you guys think about using the GPT 5.2 chat / non Codex model (as opposed to Claude) to plan and then switch to Codex 5.3 for execution?

I use both for different usecases. I do not care about personality as well (or try not to). but it is much easier to guide claude imo because with codex what ends up happening is that If you do not specify everything, he is not figuring it out by himself and I found myself less productive with Codex as a consequence. I am open to use it and your tweets genuenly inspired me to try it for an extended period of time but It was a very frustrating experience compared to Opus. and I did pretty big things with those (more of it on this soon)

i really don’t get why coding agents should come with “personality”. i find it nice that codex only cares about code quality and not your feelings. as it should be in a collaborator tool. it still has issues like not respecting the spirit of your ask while respecting the letter. in several months of use i can recall one interesting example when i found codex sneaky: i asked it to add gz to frida script and it implemented it from scratch, a move that raised an eyebrow. when i asked why not use a library, it said that it’s not possible in frida (probably true). but then it also admitted “one important caveat” that it only implemented the format but no compression.

@ilmoi Yeah, I think this is correct. AI reasoning works best when the reasoning has been done tons of times in the training dataset.

one interesting gauge for measuring domains where (current) AI will do / wont do well is hidden state hidden state = stuff not on paper, that you keep in the back of your mind coding = 0 hidden state. Everything is on github. AI is perfect. legal = lots of hidden state. You put stuff into contract, but you've thought through how the other party might challenge it and prepared 3 counters, as well as setup the contract for them. AI is great at 0 hidden state reasosning, but not at that with lots of hidden state. In the contract example it would fail to think through future challenges and counters and setup the contract correctly. you can think of it as AI failing in adversarial settings where everything can't be revealed at once.

@Wild_Randomness Right "shoulder" appearing more resilient than I expected too

$XAG $SLV That’s a good rock— put in that right shoulder… You can go to sleep now. I will awaken you in the spring.