I think Cursor is doing pretty well in the enterprise space. It seems much more useful than just throwing agents upon subagents on an unsuspecting task like Claude Code.

Cursor is fine, the example is about how things go out of hype in very little time. However I believe Cursor will not survive much. It is designed around a model that will not survive: that the AI "helps you writing code", and you review, and need an IDE like that. There are many developers that want an IDE and can't stand the terminal experience of Claude Code and Codex, but I don't believe most developers in the future will inspect closely the code written by the AIs, and things like Cursor will look like products designed for a transition step that is no longer here (already).

I'd venture a guess that most of the software in the world is not written from scratch but painstakingly maintained and as such, Cursor is a good fit while CC is not. Besides, if agentic coding does go off, Cursor has the customer relationship and can just offer it as an additional mode.

Whoever stands in front of the customer ultimately wins. The rest are just cost centers.

Or conceiving children, or any natural evolution as a species (as the adage goes, “ideas don’t die; people do”).

Speed of light doesn't adhere to Moore's law :) and it's made worse by the fact most everyone connects via WiFi these days and it alone adds a few ms more.

That's assuming no human would ever go near the code, and that over time it's not getting out of hand (inference time, token limits are all a thing), and that anti-patterns don't get to where the code is a logical mess which produces bugs through a webbing of specific behaviors instead of proper architecture.

However I guess that at least some of that can be mitigated by distilling out a system description and then running agents again to refactor the entire thing.

> However I guess that at least some of that can be mitigated by distilling out a system description and then running agents again to refactor the entire thing.

The problem with this is that the code is the spec. There are 1000 times more decisions made in the implementation details than are ever going to be recorded in a test suite or a spec.

The only way for that to work differently is if the spec is as complex as the code and at that level what’s the point.

With what you’re describing, every time you regenerate the whole thing you’re going to get different behavior, which is just madness.

You could argue that all the way down to machine code, but clearly at some point and in many cases, the abstraction in a language like Python and a heap of libraries is descriptive enough for you not to care what’s underneath.

The difference is that what those languages compile to is much much more stable than what is produced by running a spec through an LLM.

Python or a library might change the implementation of a sorting algorithm once in a few years. An LLM is likely to do it every time you regenerate the code.

It’s not just a matter of non-determinism either, but about how chaotic LLMs are. Compilers can produce different machine code with slightly different inputs, but it’s nothing compared to how wildly different LLM output is with very small differences in input. Adding a single word to your spec file can cause the final code to be unrecognizably different.

And that is the right assumption. Why would any humans need (or even want) to look at code any more? That’s like saying you want to go manually inspect the oil refinery every time you fill your car up with gas. Absurd.

Cars may be built by robots but they are maintained by human technicians. They need a reasonable layout and a service manual. I can’t fathom (yet) having an important codebase - a significant piece of a company’s IP - that is shut off to engineers for auditing and maintenance.

Assuming you did not use an LLM to craft your comment, I’d say “case in point”.

Whatsapp is a strong net positive. This is the world’s communication network (I’m counting US out because it is counting itself out)

WhatsApp was not developed by Meta. They just bought it. That said, I don't think Meta/FB is a net-negative, far from it. They contributed back to the community with high quality infra-level software.

Sometimes we in the tech community need to poke our heads out of our tech silos.

Once you do, you will see how much societal damage Meta has caused under Zuck's leadership.

In that regard I fully agree. My view was merely from a technical perspective.

And I agree with you. There are many great tech folks at Meta who released some great open source projects.

It's the leadership that's the problem.

Having ICs with no organization, synchronization or shared vision creates chaos, toxicity and a lot of technical debt. You can easily create negative value. ICs need direction to be successful, and well managed people are much happier in my experience than non-managed people.

Firstly, management and leadership are not the same thing. Giving direction is the job of a leader. Managers, just like anyone else, are rarely good leaders. They are more likely to give the wrong direction and vision than ICs, given that they typically also know less.

IC's do benefit from coordination, as any team might. That is management. However, having more than the absolute minimum of managers and management attached to a product invariably means an exponential decrease in efficiency.

Any team with more managers than senior ICs such as staff engineers is in trouble. That's because staff+ engineers are the people who's ACTUAL job it is to give direction, force multiplication and avoidance of local minima.

Hence, the nature of the position of manager is that it is very often unnecessary, or only intermittently useful. Therefore, a successful manager is not one who makes the product succeed, but rather someone who creates work that they themselves can and need to solve. Typically, this happens when there is a group of managers where there should be only one.

> Well managed people are much happier in my experienced

Emphasis on the well-managed. If the management actually helps the tram achieve their goals and doesn't stifle them, then great. Otherwise, you end up with bloat.

A company with only ICs (that produces ICs) is a whole lot more useful than a company with only managers.

There are many useful and successful project management companies that are an indispensable part of many industries, most notably in infrastructure projects.

Yet the thing got done. Perhaps in the age of AI, it’s about making things get done.

Not only do we fund israel, our leaders have been wasting trillions fighting wars on behalf of israel. The newest target israel want the US to take out is strangely enough - Iran...

US does not fund Israel. US has a strategic interest in Israel, just as much as it has in Germany, South Korea, Japan and many other places which host a huge US military presence. Unlike those outposts, the support to Israel is given in American military equipment.

Most Western governments fund Israel. The US funds it the most.

Where’s the data to support your extraordinary claim?

You could start with https://en.wikipedia.org/wiki/List_of_countries_supplying_ar...

You’ve given a list of countries selling weapons to Israel. If I sell you my car, I am not “funding” you.

What an exercise in utter futility.

You could also read https://www.cfr.org/articles/us-aid-israel-four-charts

These are just Google results.

Iran is made of many different ethnicities, and there were reports of Arab militants that were brought in by the regime (it’s not hard to imagine given how reliant those organizations are on Iran for support).

It’s generally not very hard to incite violence across groups in the Middle East, especially when you consider how bad the outcome might be for the losing side. Case in point, the Alawites who lost control of Syria and are now persecuted by the new government.