Alex Ellman

It's crazy that Claude doesn't know what Claude Cowork is

@dccockfoster It's Next JS except for Vue instead of React. I just chose it the first time I made a frontend because I heard vue is easier to work with. Works great for me but if you already know React well Next is probably fine

@aellman Never heard of Nuxt. Is it any good? Curious to try it out

The key to having Claude Code work well for you is to use the same stack every time and make skills to guide it to work with that stack in an opinionated way. I always use Django and Nuxt. I created a skill for each telling it how to organize files, where to put business logic and how to interact with each other. Now I can spin up webapps really fast while maintaining a consistent structure without massive files.

@dccockfoster Weren’t you gonna start a coffee brand… what happened to that

Buying a cold brew from a random coffee shop in NYC is a total gamble You either end up wired out of your mind and extremely jittery Or worse, you get something that looks, tastes and feels closer to water then coffee We need a NYC cold brew standards coalition

@HighyieldHarry This doesn’t impact New York literally at all

I want NY to succeed and I think the art of getting wealthy NYers to keep paying taxes/stay in the state is paramount. Zohran really poked the wrong bear.

@Geiger_Capital What a pussy

Ken Griffin says he's scaling down in NYC: "What the mayor of New York has made clear to us, is that we need to double down on Miami."

@hthieblot @UsmanReads Being a founding employee is almost always a horrible deal

@UsmanReads A few % of a possible billion dollar company

Let’s be real: There’s no work-life balance in early-stage startups. If a candidate asks about it in the interview, don’t hire them. 60+ hours a week or bust.

@tednotlasso can we stop calling every facet of life a "layer"

hot take: manhattan is relatively affordable outside of its main bottleneck - rent - if you're willing to forego the luxury layer its functional layer is very low-cost: public transport (big one), coffee, pizza, delis, free parks + museums + events most people want luxury tho

@haridigresses Large orgs still move much slower than startups, coding was never the issue it was approvals, decision heirarchies, etc. Looking forward to your longer post

@aellman The edge for startups used to be speed. Now everyone else has that too. And with this much noise in the market, distribution is king, which startups lack in the early days (while the others all have). Have a longer post coming on this. The math is totally upside down now.

If you're building AI B2B SaaS, you have more competitive threats than ever. IMO the things most likely to kill you, in order: 1. Nimble incumbents add your product as a feature (if there are indeed such nimble 800lb gorillas in your space) 2. Other startups race to build / copy what you're building more easily — making it so noisy and competitive that you can't get off the ground. 3. If you're selling to enterprise, then internal IT teams can build many more products like yours, more easily — by using existing or new platforms like Databricks, Palantir, etc. FWIW I don't think Lovable or Replit will be dominant options here. 4. OpenAI / Anthropic / etc. move into your space 5. Your buyer (whichever function you sell into) builds your product internally. This will happen but won't be very common. Most people at most companies do not want to build anything. Buyers like off-the-shelf stuff, and really want a done-for-you implementation with training, support, etc.

@ivanburazin They’re not going to after either

"Don't date until Series B" is a self-imposed code for folks who wouldn't get a date until Series B anyway.

@dccockfoster Alright I do agree with this

Many people in crypto are worried that TradFi is going to eat everyone’s lunch. This won’t happen bc too many TradFi people suffer from ‘this is how it’s always been done’ syndrome Crypto natives have the creativity to imagine paradigm shifts

@yourealazyfvck Love this

If you're so smart, why aren't you rich?

@dccockfoster Yeah should’ve said a primary use case. I’m just saying what it looks like from an outsider

@aellman The premise is totally wrong, in what world are prediction markets the primary use case for crypto? Stablecoins & Bitcoin are larger then prediction markets

Almost every new financial innovation gets called ‘gambling’ and shunned at first VC was widely derided in the 70’s - pension funds were banned from investing in it Today 30% of the Nasdaq was funded by VC Crypto is following a similar path to all other finance innovations

@dccockfoster I don’t think crypto is just for gambling just curious what you think about that!

@dccockfoster That’s true, but can you blame people for being skeptical when the current major use case is for prediction markets? Crypto obviously has other use cases but did VC look like that when it came out?

@dccockfoster Lmao

Bezos really did a number on the world when he came up with the phrase ‘if the anecdotes and the data disagree, the data is usually wrong’ I’ve heard so many people use it to justify completely ignoring correct and valid data based on a few irrelevant anecdotes lmao

@TanmaySaboo True but if Anthropic can assemble 20k lawyers with a simple webinar announcement it shows they already have a good amount of trust

@aellman Labs can copy features overnight. Trust, compliance, and client relationships? That takes years. Harvey's real moat was never the UI. 🧠

Anthropic assembled 20,000 lawyers yesterday to watch a full on assault by a lab on the app layer. In their 1 hour webinar Anthropic's in-house counsel showed how Claude Cowork can be used to replicate the main features of legal tech's largest startups. The crazy part is that Anthropic didn't even build a UI specific to the legal workflow. Their legal plugin is just a collection of skill files that instruct Claude to perform legal tasks like document review, legal research etc. and a Microsoft Word extension. This seriously calls into question the massive valuation legal startups are getting. Harvey's $11b valuation cannot be defended by its UI when labs can replicate their features as a side project. It is perhaps a positive for vertical startups that the top questions from the audience were not about workflows but about data security and privilege. This shows where much of the differentiation in this category will come from: trust. A lab just showed they can replicate the main features of vertical software but it is still an open question on who their prospective customers will trust with their data.

@signulll Yes, and matching Claude with a framework like ShadCDN with guidelines will help it conform to good design. The info on how to make good design is all out there and not a secret.

if taste is a learned discriminator over aesthetic objects that produce consistent, identifiable preferences, then ai has taste. more than that, it has a real aesthetic sense. you see this immediately when you interact with it the way you would a designer. ask it to critique, refine, or generate & it responds with outputs that are tasteful. a lot of ppl just try to assess one shot output but good design needs iteration.. just like anything. i’m pretty obsessive about design, & ai, particularly claude, can reliably produce work that aligns with what we’d call good taste imho. will it produce slop? absolutely. but humans also produce slop. why do we have a higher bar for ai?