Silhouette

The page above still includes similar changes to the terms but now says they come into effect on 20 November 2026.

Does anyone know what's going on here? We were about to delete our business account today because we can't risk the potential consequences of the new terms but maybe we have a few more days? Has there been more push back from the big players than has been widely reported?

I saw an interesting observation the other day that Jensen Huang has played this game the best because his company is the one actually selling real products and making real money in return. If all the AI giants are buying up huge amounts of hardware speculatively because they assume they will need those data centres in the future and then the bottom falls out of the market then it's the AI companies left holding the baby. And if we do get to a point where local models are competitive then the OpenAIs and Anthropics of the industry might face real difficulties but it would just be a different market for Nvidia to also be selling real hardware and still in exchange for real money.

If progress with frontier models doesn't continue at the same pace then the investments and funny money deals might start to look very fragile.

If progress does continue then that progress might be towards smaller models more than quantum steps up in quality or capability that would increase the risk of the scenario above.

All of this is set against a backdrop of public hostility and political problems that I can only imagine increasing in the near term as well.

This feels like an unusually high stakes game of poker. I don't think I'd want to bet on Big AI right now if I had a few billion lying around. Then again - maybe that's why I don't have a few billion lying around!

I would say that a system with some sort of binary evaluation that indicates whether it succeeded or not is receiving feedback. It could also be some kind of sliding scale - how close did it get?

Either of those might be useful for determining a good future direction to iterate in but that would surely depend on the nature of the problem and how it's possible to search for solutions. I'd still call it a feedback loop even if the evaluation wasn't enough to point in a specific direction for a further iteration though. YMMV.

A recurring problem with politics is that even a lot of decent and reasonably intelligent people still aren't interested enough to spend a lot of their time looking into political issues in detail. And that's understandable! Most adults have enough to worry about with jobs and families and their own hobbies and interests. The point of representative democracy as commonly used in Western nations is to have representatives who can devote their time to getting involved and understanding the issues on behalf of those they represent. But then how do voters know that the people they vote to represent them will take a good position on the issues that matter to them? A lot of it comes down to rhetoric and trust - and that's how folksy populist politicians manage to win votes and get elected even if better informed voters can see straight through the sales pitch.

The kind of public information campaigns I'm talking about here are the ones that try to raise awareness of the most important points about the most important political issues among the general population who can't or won't do enough research to find out for themselves. It's about getting the core message and basic facts to enough people who would then support doing sensible things as long as they have that essential information. But of course there can be a fine line between public information based on facts and logical reasoning from one party and mere political propaganda from another. The same tools of government and taxpayer funding can support either in exactly the same way - depending on who is already in power at the time.

The only way this changes is if there is real political/government leadership on the issue. That leadership can direct national resources. It can legislate. It can even tax - if necessarily to a very high degree - less useful ways of spending available resources in the private sector to deter waste.

We can't get that kind of leadership in many countries at the moment because too many people will just vote for whichever party they always have or whichever populist nutter has the most convincing sound bites. So before the political problem can be resolved there needs to be much more public education on the realities of the issue.

This is unfortunately a circular argument because the public information campaigns that are needed to fix the education problem are only going to be run by political/government leadership of the required calibre. If anything the populist lightweights will push the other way because dumbing down the voting population helps to sustain their own control and importance.

My hope is that before it's too late there will be enough politicians who are both capable of dealing with the real issues here and capable of taking on the populists convincingly enough to get elected to start moving the needle. Unfortunately I don't really see any other way this gets fixed instead of spiralling downwards.

In particular I don't see why we should assume that the capabilities of AI programming tools will continue to improve at the same pace. There is no obvious source of new and high quality data for anyone to slurp up to train the next generation of models. Much of the change over the past year hasn't come from the incremental improvements in the models anyway. In fact I've even seen a few comments online arguing that some of the newer models had regressed in some respects. It has been the shift towards using agents and automating the incremental refinement of work in progress instead of waiting for human intervention that has been the big change and I don't think we really know yet whether the problems we're already seeing with that approach can be resolved effectively.

Not any time. Your feedback loop must provide some direction so the iteration is at least heading towards an improvement. For novel applications you might not have a convenient way to do that yet.

Even with that provision there's still a possibility of incrementally updating output converging on a local maximum. That local maximum might not be a solution at all. If it is a solution then the model might still miss much better solutions that a real developer with more insight might find.

As you said yourself the models need external feedback if they don't have sufficient information to make adjustments by themselves. In some cases that feedback might be "Start again because you're hopelessly off track". Sometimes even when they start again they just get hopelessly off-track again. That can happen even if they are explicitly prompted to do something else. So even closely supervising the LLM and not using self-iterating agents isn't guaranteed to get good results for novel applications.

Would I routinely code in assembly today? Not for (almost) any new project.

Can I understand assembly and have I used that skill to understand performance problems or debug unexpected behaviour (including some that really did turn out to be caused by bugs in popular compilers)? Absolutely.

I'm not generally a fan of the comparison between what LLMs do and what compilers do. Compilers might vary in their output for given source code but they still follow (hopefully) well-defined specifications. Their behaviour is deterministic and the behaviour of their output is deterministic. With sufficient understanding you can reason about that output or diagnose faults in the compiler behaviour if you ever need to. None of this is true for LLMs.

1. We are working on different types of application and code.

2. I am in a bubble and no-one in it has experienced the standards and productivity that SOTA AI tools can achieve.

3. The advocates are in a bubble and no-one in it has experienced the standards and productivity that expert developers can achieve.

I hope the answer is 1. I fear the answer is 2. I suspect the answer is 3.

That certainly can be true but I would challenge "generally". If you're working on something novel then it's not exactly unusual for even SOTA models to produce a first attempt that makes some very bad assumptions but then get stuck when instructed to use correct axioms instead. This looks a lot like the suduko meme - without any real insight into the problem the models end up just iterating around in circles and never finding a solution.

You probably won't see this very often if you're working on simple CRUD web applications where there are millions if not billions of lines of code in the training data set. IME you will probably see it quite frequently if you're working on - for example - software to control some kind of equipment or analyse the results of some kind of experiment where the LLM might have little if any training data about the field you're operating in.

It can feel that way with all the hype but I very much doubt that is really true. I see an industry that is rapidly dividing into two camps - "move fast and break things" and "this actually needs to work". Writing everything with agents and merging while barely even checking what is in the PR has become standard practice in the former but it will get you laughed out of the room in the latter.

All those extra tests are churning through tokens and CI resources. These are far from free in a direct financial sense. I've also noticed some of the more recent models will try to do a lot of experimentation and ad-hoc testing by spinning up shells or running something like Python scripts direct from a command line. Every time they do that you either have to review whatever they propose to run carefully - which is slow and tedious work - or stick the tools in YOLO mode and hope that your agents don't accidentally wipe your entire filesystem, open your firewall for malware to take over your network, and perhaps launch a few nuclear missiles.

I would respectfully disagree about the value of unit tests as well. The reason unit tests appear to have less value in the LLM/agent era is because current LLMs suck at decomposition and other architecture-level work. They just see a mass of code and change it as a mass of code. This probably results in orders of magnitude more code being written/changed than would have been necessary for a skilled developer working on a well-maintained code base to solve the original problem and that automated work also takes longer in wall time. Again this gets brushed under the carpet because a lot of people trust that LLMs are so good that their costs are negligible regardless of how much work they do. People used to say that about the Cloud transition as well because not needing so many in-house staff to manage your resources was going to be a huge saving. Today we have companies bringing in consultants because they literally can't even work out how much they're spending on AWS by themselves. Instead of having a few greybeards who could make a Linux server do anything we have a few increasinglygreybeards who know which combination of 17 AWS resources they need Terraform to spin up so they can write the web server equivalent of hello-world.c. Meet the new boss - same as the old boss. :)