idk who needs to hear this but stop trying to get into film just buy a fujifilm camera it's cute to tote around a vintage nikon. it's not cute to spend $2 or $3 a shot when you're just getting into photography get a fujifilm. it doesn't matter which one, if you like it you can upgrade later x100 x-t5 x-t50 target $1,000 for camera lens. sounds like a lot but that's only 300 shots on a film camera. 10 rolls. add a few film simulation recipes. start with "Reggie's Portra" (google it). follow the steps. punch it into the camera. set it to auto and just blast away. you'll see immediately that it's vastly superior to the iPhone camera fujis are especially good for people photos, the color rendering on skin is magical your wife will especially appreciate this purchase. this is an approved hobby good luck

New York has 2021 summer energy. COVID is over energy. The euphoria of being alive energy. If you don’t meet your soul mate on the street by Labor Day this year I just can’t help you.

huge missed opportunity to name spacex rocket internet

u eventually start to love the game for the game. u wake up and do the work for the sake of the work and nothing else. To labor in arts or entrepreneurship for any reason other than love is prostitution

Quality of yap is the only durable correlate to a long relationship. Everything else undergoes hedonic adaptation: looks, money, status. Yapping with the right person never gets old.

funemployment used to mean backpacking across Asia for six months. now it means building 5 vibe coded apps that generate zero revenue

The rental situation in New York is truly catastrophic. In Fort Greene, Prospect Heights, and Clinton hill right now there are 94 2-bedroom rentals available at any price. 85,000 people live here and there are <100 units on market that could reasonably fit a family.

schwab is literally advertising bitcoin and eth to me on the homepage--do they not spend time on crypto twitter? do they not understand that the meta is now revenue, privacy, AI, and hyperliquid...? don't they understand bitcoin quantum risks and lack of eth foundation alignment to improving eth price? surely they know eth is inflationary and bitcoin has no more narrative tailwinds especially as an inflation hedge. do they have no understanding of market dispersion? crypto will be successful in spite of bitcoin and eth not because of them. real use-cases are coming. blockchains, not crypto. zcash.

Rotated a big chunk of S&P into SaaS names in February and March and suffered a few months of bleedout. Had doubts for a bit but now the whole sector's rerated. $IGV up ~35% from the April bottom. I think there is further rerating to come when OAI and Anthropic S-1's drop. By then the AI industrial complex has to have proven unit economics ROI to adopters. Not clear they've done either yet. The good vertical software players actually have data and expertise that are strengthened by AI because it improves the interface for customers. So when people realize vertical software doesn't get killed by AI, multiples expand again. The labs are ultimately selling commodity tokens which don't carry expertise, aren't attached to proprietary data and are easy to replace with Chinese alternatives. As models get better and better, it becomes easier to replace workflows with efficient local models or open source alternatives. This is obvious. There is massive FUD from enterprise buyers on the timeline about AI adoption productivity gains that has to be answered before these companies go public. And outside of Cognition putting their nuts on the line and guaranteeing productivity gains in engineering deployments (the one place where ROI should be easy to prove), I'm not sure I see the FUD slowing. Current thinking is rotate more S&P into software and relative value Mag 7 hyperscalers like Google and Amazon. If AI as a bet doesn't work out, these guys have insanely profitable franchises growing 30% at scale, plus they turn off capex, generate even more cash and move onto the next kooky moonshot to bet on. But in a capex slodown NVIDIA gets absolutely murdered, not to mention the pureplay hyperscalers like Coreweave. Not happening anytime soon, so not really buying a catalyst here, just a tiny bit of protection against overpriced structurally unprofitable AI names entering my happy little garden of passive indices.

"AI notepad" is a functional, commodity position which undersells the enduring magic of granola when i start a new meeting nowadays, i have multiple apps blitzing my screen to record me: > attio pops up and says "i'll enrich your CRM entries" > notion and gemini say "you already work here, keep your context here too" and granola says, "and uh, i'm here too!" over time, the granola has been reduced to my preferred backup because--on functional grounds--it *is* nicer to keep my context where it already lives, in my CRM or knowledge management tools! so by functional benefit, attio is my CRM that happens to record, and granola is my backup. and i'm happy to pay whatever i pay for granola but no more, because it's cheap context insurance and that not only erodes the monolithicness of granola's context but its ~grandeur~ as an idea -- granola at its best is an opinion on *how work should be done in the age of AI* @lee94josh has convinced me granola should be a dominant cultural brand because granola is a counterposition to undifferentiated surveillance-maxxing context aggregation it's an objection to the cold logic that our role on this earth is creating structured inputs for AI agents instead, granola is a way to stay in touch with our intuition and intelligence. note-taking is thoughtful, reflective, and human-- which is the only way you will actually compete anyway! every other recorder screams: "give me what's in your fucking head already so i can fire you and feed your lifeless context into the gaping maw of an all-knowing machine intelligence!!" so as everyone furiously churns the whitewater of agentic workflows, granola says: what if meetings were unintrusive? intentional. thoughtful. present. and what if as a result, you were *more* productive, not less? -- so "AI notepad" is...fine i guess. sometimes i take notes in granola! that *is* a functional differentiation. but granola is in competition with far stronger context systems. no matter how good granola's integrations, it doesn't have the right to suggest CRM edits or new documents, or e-mail drafts. i don't trust it to do so, because it's not attio. it's not notion. it's not gmail. so granola can compete on functions and be my backup call recorder, or-- it can represent a way of working that makes you happier, more engaged, and ultimately more productive.

AI psychosis is just model collapse from post-training on synthetic data

Bro last night was a papal encyclical

you don't have to know more than everyone else you just have to know more than your customer

toddler loves being hunted like prey she can't really walk yet, so i have my wife hold her over her shoulder while i stalk through an imaginary forest, leaping out at random times to tickle attack her she loves it. kicking, screaming, cackling why is this

~1 week into letting it rip on twitter: 1. this is fun 2. distribution is more bimodal than i thought. the vast majority of tweets are <10k views, <100 likes, <10 comments, but 1 or 2 this week really ran 3. it's easy to tell when something's going to run 4. the joy of creation is highly correlated with performance on X. you can't instrumentalize twitter as a means, you have to see joy in the act itself 5. "relatability" might be more powerful than credibility. it's very different here than on LinkedIn. twitter cares about proximity to the human experience; linkedin cares about proximity to status 6. write everything down. if you have the spark of an idea, write it down and let it take seed in your mind. over time, it will germinate into something worth sharing. this is an iterative process that is not shortcuttable by simply putting your idea into AI and hoping it closes the loop for you. your subconscious will do work to solve the problems you lay at its feet. before you give up on something, add the ingredient of time and let it ferment.

the simple explanation for why AI sucks at tweeting is because AI sucks at writing, and twitter users are discerning and the reason why it's tolerated on linkedin is because of the linkedin social contract: - we are all trying to social climb our way to the top - let's not burn bridges - so everyone be nice and pretend to like each other dunking on anything is uncouth. linkedin is a linear algorithm that still creates distribution via followings and shares, so network determinism is high twitter which has much greater discoverability and virality, but also more volatility. it's far more likely for you to post something on twitter and for it to spread into a circle on the internet that doesn't contain your friends. therefore on linkedin, social status matters >>> content quality (as dell's COO showed below) LLMs are aesthetically convergent, which survives well within domains prizing compliance and in-group affiliation (linkedin) twitter's pseudanon culture-algorithm rewards honesty divergence so twitter is more interested in truthiness than association

we can all make pinanfarina slop. we can all take the model trained on a million ferrari photos and make an image that looks like a ferrari. we can all one-shot a better looking car than what jony ive just came up with. i get the logic. look here, with close to zero effort i can conjure a thing of objective beauty. but with even less effort than that, you can simply pull up an old photo of a 250 GTO and say, "this is beautiful." there's something deeply insidious about LLM "creation": the danger the luce portends, which i don't see abating anytime soon--is that in the domains in which we cannot create, have no notion of what it takes to create, creation now seems within our grasp. we have been granted a tool whose output is by definition maximally derivative, yet convinces us we're each an inventor

Our breathless obsession with curators and tastemakers could be turned a few degrees and pointed toward a romance with editors. "But editors are still the world's readers. And thus the eyes of the world," wrote the editor Betsy Lerner. An editor is a writer's first and most important reader. She has the utmost empathy for everyone who comes after her. She inhabits the trusted station of confronting the writer with a reader's true feelings. Joy, anguish, pain, patience, annoyance. The most painful thing about editing is killing the beautiful, invasive sentences who crawl into the garden of your writing. Seductive weeds of words who choke the sun and starve the soil. A muse inspires. An editor conspires. Arthur Quiller-Couch said, "Murder your darlings." An editor is the one who cups your hands as you thrust the knife. This is a love letter to my editor Marisa Samek. Thank you for helping me murder my darlings.

if you're a community-driven business i.e. your GTM motion depends on people sharing outputs from your product you need to be PUMPING use-case testimonials from your power users 1. capture e-mails at signup 2. enrich with Clay 3. feed to Claude and categorize by credibility according to a scorecard you determine (most of the time this is just title company, but it could be e.g. social media followers for some products) 4. AS THE FOUNDER, hand-message the user and request an interview. if your product data shows they are a power user / superfan, this should have an insanely high close rate 5. ask to record the interview, then simply ask about the biggest wins and insights they've had using the product 6. ask for a headshot and permission to publish quotes 7. do 1 of these a week, every week in a many-to-many business, the user is not learning from you, they are learning *from other users,* so the social proof isn't "logos," it's use-cases. and a lot more businesses than you'd think are community-driven. AI coding tools, imagegen, even many (most) enterprise pro tools tl;dr - who are the legit-ass people using your product - what amazing things are they doing with it go forth

people don't want agents, they want embedded experts "agents" are just the AI word for automation RPA (robotic process automation) has been around forever. if you have a bunch of buttons you have to click over and over again, you go, "i tire of this. please, robot, click these buttons for me" and you trust it because you have clicked the buttons many times yourself when you have a really good sense of what needs to be done, automation is great! when you don't quite have a sense of what needs to be done, automation is at best wasteful and at worst dangerous the agents that work have to work on top of a trustworthy foundational substrate: a) a deep, fact-based understanding of your business ("what is the situation at hand") b) embedded expertise ("what is the empirically best way to do this") not sure how this is done without human touch at first. that's why the pure-agent startups ("we'll just plug into your data and replace your CMO!") are not only unbelievable, but indefensible they will run into detractors who will rightly say, "why can't claude do this?" the model for incredible, sticky services-as-software businesses looks like: 1) initial low-margin, high-value consulting and implementation 2) delivering high velocity of business actions via agents with embedded human expertise