peterthehacker

I don’t think this model is bootstrap-able. It’s pretty much a social media play, which requires lots of burn to acquire first cohort of users.

My friend bootstrapped a genius for books competitor with a subscription model and it was really difficult to grow. It’s hard to build value based on network effects with subscription model.

It sounds like a Herculean marketing effort to rebrand from rap music to all music then to sophisticated literature. Those markets are completely different.

Then of course theres the copyright issues.

> Eventually the cloud is going to burst and everyone’s data will be public

Why? This was Epik being hacked not AWS or Azure. It’s just a domain registrar. And a shady one at that. Their lack of security is not indicative of the rest of the cloud.

Do governments have the technologies to become carbon neutral?

From my maybe uniformed perspective it seems like we’re still several scientific breakthroughs away from alternative energies and sustainable technologies being competitive and cost effective.

Only half of California homes are owner occupied. [0] So, there could be more CA voters renting than home owners. To put that in perspective, Texas has over 60% owner-occupied housing [1].

[0]https://www.census.gov/quickfacts/CA

[1]https://www.census.gov/quickfacts/TX

The sample size is in the article:

> Nearly 13,500 drivers were randomly sent one of two policy review forms to sign…

The distribution referenced is of mileage, which you’d expect to have some kind of right-skewed, continuous distribution.

While you’re right in a broad sense your points are irrelevant here. This is about fraudulent data (apparently created with a random number generator) being used. This is not and should not be something we expect from scientists. If this story was solely about the replication crisis in behavioral economics, then your comment would be relevant.

> In the first sign of something amiss, the 13,488 drivers in the study reported equally distributed levels of driving over the period of time covered in the study. In other words, just as many people racked up 500 miles as those who drove 10,000 miles as 40,000-milers. Also, not a single one went over 50,000.

I have a hard time believing that Dan Ariely didn’t know about this. The uniform distribution of mileage makes no sense, so this should’ve been caught right away. Plotting a histogram of the mileage data would’ve been one of the first things Ariely’s team did with this data.

That chart with “Weak Foundation” vs “Solid Foundation” reminds me of Martin Fowler’s chart in his Refactoring book that compares “good design” vs “no design”[0].

[0] https://martinfowler.com/bliki/DesignStaminaHypothesis.html

To be clear: staying in business is the universal success metric I mentioned and cited earlier (not funding or exit). Here’s another source citing the same 9/10 failure (10% success) rate for all startups in the U.S.

> Many small businesses start up every month but the failure rate is high. As of 2019, startup failure rates are around 90%. 21.5% of startups fail in the first year, 30% in the second year, 50% in the fifth year, and 70% in their 10th year. [0]

Small bootstrapped companies can still exist and succeed against these odds (several are cited in the article), but to make the claim that these odds don’t apply to bootstrapped startups is foolish wishful thinking.

[0] https://www.investopedia.com/articles/personal-finance/04091...

That’s not survivorship bias. And the measures of success in the article I cited are universal. The way you’re describing success (not going out of business) is included in those sources.

But your example, Screenhero clones like Pop (https://pop.com/), are in a large, competitive market. That doesn’t fit the qualifications of a small, underserved niche like the article describes.

How is that survivorship bias?

Bootstrapped, small, one person companies are still startups, where the odds of success are less than 10%, while venture-backed startups is 20%-30%. [0][1] What non-survivors are excluded from these cohorts?

[0] https://www.wsj.com/articles/SB10000872396390443720204578004... [1] https://www.embroker.com/blog/startup-statistics/

This has been analyzed by many, but the stats you see reported often are that 9/10 startups fail while 7/10 venture-backed startups fail [0]. Perhaps the reason why a startup has no investor interest is because the market is too small, so cutting the data by TAM (if possible) might be interesting. We don’t have data for these micro-TAM startups showing whether they succeed more or less than any other startup. But the data we do have does show an advantage to raising money from VC vs. not. But to be clear, my point is that funding should be a more complicated and nuanced discussion. And if there’s a $5M+ potential return, like the author proposes, then there should be investor interest. The question is whether it’s a good idea for the company to raise.

[0] https://review42.com/resources/what-percentage-of-startups-f...

Finding a market like that sounds really difficult to do. How does someone discover a market like this without a ton of luck?

> In this universe, of bootstrapping to several hundred thousand per month in revenue, a founder can walk away with $5m, $10m, $20m (depending on the multiple) in around half the time it’d take to build a “unicorn”.

The odds of this level of success for a solo bootstrapped founder is much smaller when you don’t raise. If you look at the stats on startup success (exits), funding has a strong positive correlation.

There are good and bad reasons to raise capital and many startups do raise for bad reasons, but that doesn’t mean that funding is inherently bad or should be eliminated as an option.

Can Judo build native desktop apps as well or just mobile apps?

Whenever I see a business try “co-leader” stuff I think of that episode of the office where Jim and Michael are promoted to “co-managers”, then Oscar comments, dripping with sarcasm:

> Look, it doesn’t take a genius to know that any organization thrives when it has two leaders. Go ahead, name a country that doesn’t have two presidents, a boat that sets sail without two captains. Where would Catholicism be without the popes?

https://m.youtube.com/watch?v=wF8jmFoQ7zM

Does Tesla have data showing how much better or worse AutoPilot is at driving vs a human driver?

> Given this, why does Copilot write such crappy code?

> The reason is because of how language models work. They show how, on average, most people write. They don’t have any sense of what’s correct or what’s good. Most code on GitHub is (by software standards) pretty old, and (by definition) written by average programmers.

Does Copilot apply any project quality filters before training? A small minority of public repos on GitHub are awesome projects that are popular and maintained, while the rest are forgotten experiments, side projects and interview answers.

Good point. I was thinking of changing styles via DOM, like CSS-in-JS, but you’re right that “y to CSS” would still be there.

It definitely adds to the bundle size, but how much it adds depends on your implementation. This discussion on the topic is pretty interesting [0]. One comment from that discussion says “it is possible to double to the size of your source code”.

[0] https://dev.to/haruanm/does-typescript-increase-the-bundle-s...

Technologies like TypeScript are still dependencies that require transpiling to vanilla JavaScript and add to the bundle size. While typescript is becoming more and more common (for good reason) it still is effectively a workaround that compensates for the unergonomic-ness of the browser environment. You could make the same argument for frameworks like Angular, React and Vue.

If WebAssembly replaced js, then newer, more efficient browser languages could be invented. Like a new implementation of typescript could be created that is statically compiled and more efficient. All of this transpiling from x to js and y to css would be unnecessary.

This project will transpile python code to JavaScript code instead of something WASM-based, like pyodide [0].

What are there the performance and usability consequences of each approach (py to js vs. py to wasm)?

[0] https://github.com/pyodide/pyodide

I stumbled upon a security issue with your site. I was reading this LP trying to understand what this product does and saw this quote

> Military-grade encryption and security

Then I checked your site’s CSP policy on Mozilla Observatory:

https://observatory.mozilla.org/analyze/coinrule.com

coinrule.com scored 0/100 (F), because your site does not have the Content Security header implemented. I’d expect a software product focusing on security to have a strong CSP.

I believe this is the essay you’re referring to:

Before the Startup: http://www.paulgraham.com/before.html

This title is misleading to me. It’s not about “best practices” but instead about “best tools”.

There’s no mention of kubernetes persistent volume claims.. it does mention that you can swap out a DATABASE_URL env var here:

> Or perhaps you want to use a managed PostgreSQL database in production but run PostgreSQL locally in a container for development.

It’s mentioned in passing with no discussion of why you should not run your database in docker compose - a very common use case - so the reasons I mentioned were not explicitly discussed.

> Snowflake will eventually fall to something else due it's poor economics.

Can you elaborate on this point? What’s wrong with their model?

There are integration companies that solve this specific use-case. I’ve used Fivetran [0] and highly recommend it. They will extract-load data from your SaaS to your warehouse and your data scientists can run SQL against the tables. Their most popular warehouses are Redshift and Snowflake. So you can still use a centralized data warehouse without dedicating internal resources to the integrations.

[0] https://fivetran.com/