For returning customers, a customer with an app installed is simply going to have a lot less friction to open the installed app and place an order than a customer who has to open their browser, log in again, and then place the order.

That alone is worth the investment for many companies in making an app, even if a minority of customers actually choose to install the app and keep it installed.

Karpathy clearly wrote something different that what you say he wrote.

> If you need to introduce a new pattern, consider refactoring the entire codebase to use this new pattern, rather than creating one-off inconsistencies.

Putting aside the mis-application of "pattern" (which _should_ be used with respect to a specific design problem, per the Gang of Four), this suggestion to "refactor the entire codebase" is impractical and calcifying.

Consistency increases legibility, but only to a certain point. If the problems that your software is trying to solve drift (as they always do with successful software), the solutions that your software employs must also shift accordingly. You can do this gradually, experimenting with possible new solutions and implementations and patterns, as you get a feel for the new problems you are solving, or you can insist on "consistency" and then find yourself having to perform a Big Rewrite under unrealistic pressure.

Safety research on toy models will continue to provide developments, but the industry expectation appears to be that emergent properties puts a low ceiling on what can be learned about safety without researching on cutting edge models.

Altman touted the governance structure of OpenAI as a mechanism for ensuring the organisation's prioritisation of safety, but the reports of internal reallocation away from safety towards keeping ChatGPT running under load concern me. Now the board has demonstrated that it was technically capable but insufficiently powerful to keep these interests in line, it seems unclear how any safety-oriented organisation, including Anthropic, could avoid the accelerationist influence of funders.

As others have pointed out, these licences appear unenforceable.

The publisher simply wants to appear responsible. There are likely many open-source-oriented engineers and scientists at these tech firms who have been pushing for publication. (See the discussion between Mark Zuckerberg and Lex Fridman.) The involved tech firms only care about the licence as far as it might minimise the likelihood of public backlash if any of these published models cause harm.