sealeck

I think it's a bit hard to say that this is definitively true: people have always been interested in running linear algebra on computers. In the absence of NVIDIA some other company would likely have found a different industry and sold linear algebra processing hardware to them!

Even when you build cool things it's respectful not to plant them in HN comments :)

I think the usual solution to this is to talk about cool stuff you've done that is only incidentally relevant to the product you're selling. For example, some detail on how you built a technical system or solved a problem, etc...

Converge towards what though... I think the level of testing/verification you need to have an LLM output a non-trivial feature (e.g. Paxos/anything with concurrency, business logic that isn't just "fetch value from spreadsheet, add to another number and save to the database") is pretty high.

Go is an implementation detail for Kubernetes and Docker...

Are you an NVIDIA fanboy?

This is a _remarkably_ aggressive comment!

Apple is selling Swift as a product - it is their preferred interface for constructing iOS applications!

Go has a critical mass that Swift clearly doesn't (i.e. there are many, many companies who have net profits of >$1bn and write most of their server software in Go).

Additionally Google isn't selling Go as a product in the same way as Apple does Swift (and where Google does publish public Go APIs it also tends to use them in the same way as their users do, so the interests are more aligned)...

> Rust wasn’t designed for any specific platform

I suspect that Mozilla being the primary developer and sponsor for many years actually meant that compatibility with all major platforms was prioritised; Mozilla obviously care about stuff working on Windows, and run lots of builds on Windows + I imagine a number of Firefox developers (if not drive) at least own a Windows machine for testing Windows-specific stuff!

I call out Windows because I think generally software people go for Mac > Linux > Windows (although Mac > Linux may be slowly changing due to liquid glass).

If the median UK salary is >£35,000 I really wonder how arrive at the conclusion that missing a flight will set you back "years or decades"...

Clever

I actually don't think this is true, and certainly of people who cover LLMs Simon Willison is one of the more critical and measured people.

It's not just about performance: having an in-process MVCC engine would simplify the implementation of many single-machine concurrent applications. Currently you usually have to combine SQLite with some kind of concurrency primitives; this is extremely painful because most OS-level concurrency primitives are really easy to misuse (e.g. it's trivial to accidentally add deadlocks, and very hard to spot and remove these ahead of time: example hard to spot concurrency bugs https://fly.io/blog/corrosion/, https://rfd.shared.oxide.computer/rfd/400)

I'd love to see what happens if you hook this renderer up to AFL++...

I thought the point of Turso was to offer better concurrency than SQLite currently does. A la https://turso.tech/blog/beyond-the-single-writer-limitation-... and https://penberg.org/papers/penberg-edgesys24.pdf

Would be great if one of the Turso developers can clarify :)

Why is writing code to execute a program using the fewest instructions possible on a virtual machine a waste of time?

The assumption that treasuries is safe depends upon the guarantor (the US government) being reliable. Currently this is ahistorically far from being the case and therefore people are investigating other places to park capital.

I think the other question is how far away this is from a "working" browser. It isn't impossible to render a meaningful subset of HTML (especially when you use external libraries to handle a lot of this). The real difficulty is doing this (a) quickly, (b) correctly and (c) securely. All of those are very hard problems, and also quite tricky to verify.

I think this kind of approach is interesting, but it's a bit sad that Cursor didn't discuss how they close the feedback loop: testing/verification. As generating code becomes cheaper, I think effort will shift to how we can more cheaply and reliably determine whether an arbitrary piece of code meets a desired specification. For example did they use https://web-platform-tests.org/, fuzz testing (e.g. feed in random webpages and inform the LLM when the fuzzer finds crashes), etc? I would imagine truly scaling long-running autonomous coding would have an emphasis on this.

Of course Cursor may well have done this, but it wasn't super deeply discussed in their blog post.

I really enjoy reading your blog and it would be super cool to see you look at approaches people have to ensuring that LLM-produced code is reliable/correct.

This is a bit like saying everyone would be a bit less jaded if the plane staying in the air wasn't hung over the Boeing 737 MAX 8 designer's heads by certain communities and used as an existential threat to the company.

I don't think many (if any) programmers would imagine 100M lines of code to mean 100,000 lines of code and not 1,000,000...

What was your dissertation, and how would Nano-Banana-pro with Veo 3.1 have helped it?

> because a "bazooka" is easy defeated by the same ratchet mechanism

That's an argument for capitulation in general: it's not an argument specifically against extending the field of scope to include services.

Why isn't AI just another application that can be run on the device? Surely we expose the necessary interfaces through the OS and the application goes from there?

As is "Donald Trump ordered the demolition of the existing east wing of the White House, in order to construct a new ballroom on that site. Microsoft has funded this construction work."

> Startups belong to the latter.

Except that Apple, Intel, Tesla, etc have all received US government investment [1]. TSMC is a product of the Taiwanese state! Government investment can be done well, and seeds excellent companies.

[1]: https://www.sba.gov/blog/2024/2024-02/white-house-sba-announ...

This sounds like a terrible approach to accounting. Surely large public companies should account for their expenditure at greater fidelity than billions.

Glad that every statement of personal opinion is now "propaganda"...

You do know you can have high availability without using cloud providers? E.g. you run a second server in a different database as a standby that can take over, etc.

In the previous post on the Fly.io blog (https://fly.io/blog/litestream-revamped/), a really cool thing was mentioned!

> In Litestream, we’re solving the problem a different way. Modern object stores like S3 and Tigris solve this problem for us: they now offer conditional write support. With conditional writes, we can implement a time-based lease. We get essentially the same constraint Consul gave us, but without having to think about it or set up a dependency.

Reading this blog post though, I couldn't see any reference to this. Is this supported in Litestream v0.5.0, or will it be for a future release?

> I think OpenAI is actually doing a great job at easing people into these new technologies. It's not such a huge leap in capabilities that it's shocking, and it helps people acclimate for what's coming. This version is still limited but you can tell that in another generation or two it's going to break through some major capabilities threshold.

This is a truly _wild_ way to describe "this version isn't much better than the previous one". Would you say "Apple's latest iPhone is a pretty small marginal improvement over the previous one, but it's useful to help peopel to acclimate for what's coming".

> Of course Dropbox is dumb when rsync exists. Of course og Twitter was dumb when group sms existed. Of course Bitcoin is dumb because...(waves hands in all encompassing disbelief).

Twitter/Dropbox offer more convenience to people: Bitcoin offers _less_. That's why people are using Bitcoin for two things: crime and speculation. People aren't buying houses with Bitcoin, they don't use it as a store of value (because it has extreme deflation), and they aren't doing their books using BTC as the currency.

What features/improvements does Bitcoin offer over the financial system? So far, the differences to me seem to be

- It's easy (or at least possible) to get a bank transaction reversed by going to court.

- If you forget your password, your bank can reset it and you don't lose all your money.

- When you transfer money using your bank, it arrives immediately in the recipient's account. When you transfer money using Bitcoin, it takes hours. The bank checks that the name matches the bank account number beforehand.

- If you get hacked, and the hacker tries to steal all your money, your bank's fraud detection system will kick in and stop this.

- Banks are regulated, and the currency is stable. This means that you can quote a price in USD, and do your accounting in USD, whereas very few people do their accounting in BTC.

- People have faith in fiat currencies because they have state backing, which means that the central bank will intervene to maintain price stability.