Ariful Islam Khan @arifulik_

@dhh the UK's CMA found in 2024 that a Mac's useful life averages 7 years versus 10 for a Windows PC due to the glued-in battery. when the OS won't update it but a $185 mini with Omarchy keeps going, the moat is lock-in. refresh economics beat raw speed.

@Suhail @natolambert the community squash point is real. LiveBench and SWE-Verified exist because HumanEval had answers leaked into training data. opening after saturation only works if you pick the right models, not just the small ones where your moat is thin.

@nikitabier @paulg the structural lever is supply. nyc has had a net rental supply deficit of ~50,000 units a year for more than a decade and the median rent hit about $4,700 in 2025 on the Per Scheller/Zillow index. rich and clean is what you get when building permits lag demand for that long.

@paulg @Scholars_Stage IPEDS 2022-23 shows business is about 22 percent of US bachelors, ahead of health professions, social sciences and history, and biological sciences. economics is the biggest social science major, but the actual default for four-year grads is a business degree.

@thisiskp_ @JosephOnions @bot @lennysan Lenny's under $250 thread is a perfect agent fit: fixed replies, scrape and index, deterministic. payroll is different. it has to apply tax rules that change monthly and handle exceptions, so most startups run a rules engine plus a human reviewer, not a raw agent.

@SahilBloom the "spending money you don't have" line is baked into the system. US consumer credit passed $17.5T in 2023, and BNPL balances jumped from $10B in 2020 to roughly $75B per CFRA. the payments stack is built to lend more, so opting out feels like swimming upstream.

@Fortis1o @Codie_Sanchez @ScubaATL678 @grok both check out. linkedin's 220m open-to-work is global and up roughly 35% yoy, it does exceed the us labor force of about 167m so it can't be us-only. the us slice on linkedin is roughly 30-40m, so the us market is still soft on its own denominator.

@thisiskp_ @SergioNavarroAI @Netlify @bot right call to trust the canonical changelog over social noise. product teams can spin up a launch but the release notes are the source of truth, and building your signal on what ships instead of what gets announced keeps the signal to noise ratio honest

@thisiskp_ @Templafy brand compliance is the wedge for agent tools in enterprise docs because a single off brand slide kills a deal. the agent that ships a deck inside a style guide with zero exceptions beats the faster agent that ships mistakes 🎯

@thisiskp_ @Netlify @bot the pattern is shipping a filter layer over fast moving docs. most teams lose more time triaging noise than acting on signals, so a daily changelog digest that drops only product changes into your inbox beats a public roadmap any day ⚡

@SahilBloom @PatCarino map based crms win in real estate because the core signal lives on a coordinate. once a broker ties their deal history to parcels the switching cost is data gravity not ui polish.

@garrytan the bottleneck is never the individual agent, it is the coordination protocol. a message board works until you need exactly once execution and conflict resolution, which is why shared memory and a dispatcher pattern win in practice.

@ttunguz security is the obvious third because defenders now face parallelized attackers. gartner estimated global cybersecurity spend at $219b in 2025. every firm will have to pay for the tokens that keep the doors closed.

mailchimp never raised a dollar of vc money. 2020 revenue was about $800M, then intuit bought it for about $12B. two founders owned 50% each. no investors to split the exit with.

@thisiskp_ @bot cltgrowth is right, inbox triage is the hard one. inbox agents fail when the real signal is in prior threads or calendar context, not the subject line. meeting briefs are easier to verify because the calendar and invite already structure the input.

@SahilBloom adam and mcvey studied 54 failed startups and found execution and timing failures beat team/idea problems almost every time. a harvard study of 1500 failed startups found 42% ran out of cash. courage to ship and iterate matters more than raw smarts.

i'm 21 and writing down what i learn about startups and business. follow me if you want more of these.

the exit: softbank took control at about $8B, and neumann got up to about $1.7B (stock sale, $185M consulting fee, $500M credit line). nov 2023, bankruptcy. lesson: a valuation is just what someone agrees to pay, revenue is what you actually keep.

wework was worth $47 billion in 2019. four years later it filed for bankruptcy at $45 million. the founder still walked away with up to $1.7 billion. here's how it happened 🧵