knivets

Don't worry, it's always ups and downs.

These new AI trends are very tiresome, very similar to 2021 crypto mania - both trigger a lot of FOMO. If we have loops that write code and we don't need to verify anything, why are the devs still here? What's point of even learning this new trick as a dev if you truly believe that this can be used without any intervention? If loops work then it follows that a loop of loop works too - why hire any people at all? Just run a bunch of loops and build a profitable business, but then what's your moat? Any person can now launch loops on top of loops.

He has early access to anthropic models, of course he will hype them up, so that they will keep sharing access to preview models with him (and more traffic to his website). It also does't require him to perform any rigorous analysis of model performance, just share how it feels:

But it's all vibes, if you want a more scientific comparison you'll have to look elsewhere.

Software engineering. During early testing, Stripe reported that Fable 5 compressed months of engineering into days. In a 50-million-line Ruby codebase, the model performed a codebase-wide migration in a day that would otherwise have taken a whole team over two months by hand.

How was it measured? How was the output of this magnitude verified over a period of couple of days?

Over the past six months, there hasn’t been a single day where I’ve checked the HN Best RSS feed without seeing a post about how AI “writes bad code,” “introduces bugs,” “creates technical debt,” or something along those lines.

because it's true

Users don’t care whether the code was written by AI or by hand, or which framework you used. They care that the product works.

How can you guarantee that it works though? You can verify, but it would be at the same speed as before the AI, or even slower.

By then, enough real-world feedback would have surfaced to identify the major issues, and tools like Claude Code would make it possible to fix and ship version 2.0 at an incredible pace.

By then you have a blackbox of a codebase which is unmaintainable, or in a worst case scenario you end up losing your data or get hacked or both.

This is not a serious analysis. No mention of open source LLMs and their impact on american AI companies. There’s also no evidence that LLMs can make significant scientific progress on their own.

Some big names in AI made predictions by pulling random dates based on vibes, the author collected this and called this data.

how long until i stop seeing this nonsense shoveled at me from every direction

Stack Overflow provides marginal value when there are LLMs. Great technical books on the other hand still provide tremendous value and complement LLMs in a learning process: a book provides a fact checked, curated set of topics with clear start and finish (a structure) and LLMs help with any blockers or missing context that readers will encounter.

This is astrology for devs.

Bot account - 70 days old, no submissions, all comments are hyping AI

easier to build your own solution than to install an existing one

seriously?

how do you know the artifact is correct?

Practically, if you have AppleCare you don't need to worry about repairability.

Refreshing to read an article with an actual engineering work as opposed to another article about AI. Great work, very inspiring!

What if your screen breaks or logic board? Top of the line MacBooks cost ~4-5k. I recently had to service a battery and they replaced a top case and a keyboard free of charge. I will continue paying for AppleCare as long as they will allow me

this is not their data though

The closer the bubble to popping the more desperate these people sound.

100% of today’s SWE tasks are done by the models.

Maybe that’s why the software is so shitty nowadays.

I have been building a code from phone web app and doogfooding a lot - https://x.com/knivets/status/2003023386080092235?s=46

Somehow this article is no longer on the first page[0] (or even second) of hn even though it has more upvotes (and is newer) than other articles with less upvotes (or older ones). Is HN hiding politically controversial articles?

[0] https://imgur.com/a/e7EplV6

Cryptocurrencies create value by creating a financial system that solves some problems that traditional finance can't. For instance, a person in some underdeveloped country can accept a payment from US, however, if there was no Bitcoin then the person would have to rely on SWIFT network (which the country might not be a part of, or under sanctions like Iran). Or just a fact that you can't be randomly locked out from your account due to cancelling or any other reason. Buying a portion of Bitcoin is like buying shares of a bank which provides a service of accepting, sending and storing your money.

Ethereum creates value by providing an infra for dApps that provide financial instruments (like deposits, lending, etc) which also create value.

As JavaScript developers, the sheer amount of stuff we're expected to know is enormous, and it grows bigger every day.

A question to ponder over: expected by whom?

We need more stuff like that

Wow, I like how readable the text is. The typography in the post is beautiful.

Rails is definitely more convenient to prototype, but the magic issue is real. You never know where a variable or a function comes from by reading the source -- you'd need to run a debugger to do that (in Python you can just follow the import statements). The overall Rails architecture feels convoluted and unnecessary complicated (Railties, Engines etc). I also dislike the fact that there is no single source of truth for data in Rails. You have schema.rb, but you can't edit that directly, it is generated by running `rails db:migrate`, so one might say that migration files are a single source of truth, which is inconvenient: I need to generate a migration, then edit the migration file to add any modifications not supported by generator script and finally add accessor and validators in model file. In Django we have a model definition as a single source of truth (data model, validations). Also, once I've created (or updated) a model definition, I run `manage.py makemigrations` and all necessary migrations (which capture the current data model state + what's necessary to do to perform a database migration automatically) are created automatically. I also like the fact that in Django data integrity is enforced by default.

Though, when it comes to prototyping, I think Rails is a much more convenient option. In Rails I can launch a working CRUD app with authentication in 10 minutes, literally. In Django I have to manually create directory structure, manually specify each route mapping, create and program controllers (views), etc. Django doesn't even have a built-in authentication templates, only controllers (views), so I end up writing this boilerplate over and over again. The other thing is that the authentication requires a username and password, which feels kind of clumsy, when every other authentication relies on email and it is not very trivial to modify that (built-in admin dashboard relies on built-in authentication for example).

Does anyone know what is the model of the fingerprint scanner used in the article?

Here's the link to the actual playground: https://quantumexperience.ng.bluemix.net/qx/editor

I think, one of the main points of writing a framework in Python or another high-level language is the ease of extension in the future by the community.

This. Often, you can't find where this variable or function comes from. You read the sources only to discover that there is no trace of this variable and it is probably defined dynamically in some callback. You can only find this kind of stuff with the help of debugger. The dynamism, which allows one to create beautiful interfaces, and is often advertised as a programmer's friend, is actually the enemy of the programmer who is going to read the code afterwards.

Banking run by humans has fundamental problems as well, that's why things like bitcoin come to existence. It is just a matter of tradeoffs.

AFAIK, most of the errors happen because users mistype the address or confuse the commission fee/amount fields, etc. UI could possible introduce some validations for fee/amount issue and probably require to retype the address or something else to avoid the other issue. Another issue is fraud, but it is a unavoidable product of such system, otherwise we end up with same problems traditional banking has. Fraud check means someone might deny the transaction even if your transaction is completely legit, for example, just because you are a citizen of a blacklisted country (my experience).

It is easy to dismiss bitcoin when you are a citizen of 1st world country -- because you are insider and have full privileges, but when you are outsider, it gets very complicated and things like bitcoin give some promise.