aroman

Right, it's about how much time the human spent - the time spent by the machine itself is irrelevant. As you rightly point out: that is why we measure programmer experience by wall clock time, not CPU time :)

No, it wouldn’t. The hours in question are human experience, not that of the agent.

Sun Tzu said: if you scrape your enemy, call it training; if your enemy scrapes you, call it an attack.

You have hundreds of hours with a model that was barely even released hundreds of hours ago?

The perception of capability varies greatly between task. For my needs for example sol xhigh consistently outperforms fable xhigh.

No. I used to use Cursor, but now my workflow is that I use an inhouse CLI tool I wrote called "bud" that wraps/seeds the harnesses per-worktree, and boots a full copy of the game so each worktree can work independently. If git worktrees solve the problem of code isolation, bud solves the problem of isolating everything else. It's about 15K lines of rust, and I use it 100 times a day or so. It's sort of a layer on top of a harness like codex/claude code.

I have 10+ of these workspaces in parallel, and I context switch between them as I get blocked on things. I manage the workspaces using `herder`, which is a terrific tmux-like tool that allows me to keep those workspaces on a nixOS machine I have at home that I SSH into via tailscale, so my agents don't stop working every time I close my laptop (it also lets me leverage that machine's computing resources instead of running dozens of servers and harnesses on my poor MacBook).

Indeed, much of the scariness is how fearlessly and confidently it writes them with little regard to their actual usefulness or value. When I find it adding a lot of tests, I often say something like: "audit each test carefully, and consider whether the test is testing a meaningful boundary or is more ceremonial. delete low-value tests and add new tests to cover meaningful boundaries not exercised by the gaps you identify". Without fail, this always produces some decent results.

Having said that, in truth, I almost never read the unit tests. Before AI, we had almost none (see: several person game studio) so the tradeoff is not "AI-generated tests" vs "human written ones", it's whether we have tests at all. So, I take them for what they're worth - not much - but if it catches an extra regression before it ships every now and then, it was worth it for the price (~free).

Like before AI, the scrutiny varies with the sensitivity of the area being edited.

Simple UI change? I do an AI review, but otherwise neither read nor write the code. The models are good enough they write better UI code than me, 9 out of 10 times. Not always the more idiomatic, but usually safer and more correct.

Change to our core data plane? I might spend 2-3 times more effort reviewing it than before AI. Yes, I go more slowly than pre-AI. Many more reviews, many more angles considered, including both human and (lots of) AI review cycles.

Most code is not that critical, and AI is also scarily good at writing tests. We also spend considerably more time paying down tech debt and testing thanks to AI, now that the cost is near-zero.

Net: I spend 10-25X less time on low-risk changes. I often direct (or at least approve) the implementation approach, but I rarely read this code. I spend 2-3X more time on high-risk changes. In both cases, I never write code "by hand". Since about November, I've had no reason to actually edit code in a code editor (perhaps maybe except .env files, which we don't allow agents to edit for obvious reasons).

AI is a tool. You can use it to go fast recklessly, or you can use it to go slow with confidence. Just like before AI... the skill and art of engineering is knowing when to do which.

In terms of ability to ship? Easily tenfold. We literally ship 10 times more than before AI. This does not, however, translate into a tenfold increase in actual business success, of course :)

Claude Code fan here... Codex is very good. Sometimes better. The killer feature is price.

After 6+ months of exclusive Claude Code usage, I was begrudgingly forced to try Codex once Anthropic rejiggered their limits such that I kept maxing out my $200/mo plan in just a few days. These days I pay both $200/mo plans, and it's just about enough to get me through a week's work (small game studio - infinite code to write!)

I've been doing this for ages - you just spin up harness B as a subprocess/tool call from harness A. For example, I had a "/codex-review" claude skill for ages that did exactly that. Technically you're right it wouldn't be switching, since you're right the two ideas are at different altitudes, but I think in practice it has the same impact: within one harness, you can delegate certain tasks to certain models or harnesses.

This makes me think they really are quite capacity constrained at the moment.

I had assumed they were primarily limiting it to entice people to upgrade, but I feel like these limits are so low and so temporary (especially over July 4th weekend in the US) that people will barely get a chance to get "used to it" and then think: "man, I can't live without this, I'll pay for API pricing".

I'm not sure I follow - 614 GB/sec is pretty squarely in dGPU territory (~5070 level). External GPUs can definitely exceed that on the very high end, but it seems pretty competitive, no?

For sure, on paper - I'm curious, do you actually notice that difference in your day-to-day? I struggle to think of times in my usage of my computer where I think "this feels slow", but maybe I'm blind to it.

You're arguing against a point I did not make. I observed that Cloudflare prioritizes expanding to new products over making improvements to existing ones. I did not claim they do not improve their products.

There are numerous examples, such as Cloudflare claiming Workflows had reached "GA" status before offering a way to delete workflows... not via wrangler, not the dashboard, not the API. They eventually added this many months after declaring it "GA", with no upfront disclaimers, just the capability silently missing.

If they want to move fast and break things (and they do, as their numerous, "CEO declares an emergency" outages have evidenced) they can — but they need to be honest in their communication about the production readiness of their new products.

That is not "life and priorities": it is poor communication/judgement from an infra company, and misleading advertising at worst. I choose to believe it's the former.

I'm confused why you would replace your 16" M1 Pro with a 15" M5 Air... on which axes is that an upgrade over your M1 Pro?

What would be the bottleneck?

My hardware isn't powerful enough to try, so I'm asking out of genuine curiosity, not to push back: what do you use DS4 for? Did it replace e.g Claude Code with Opus for you, or was it replacing something else?

The point you’re making is that AI is an intrinsically bad technology, but that does not follow from this news story, which merely evidences that AI is in demand and supply has not caught up.

Who exactly is “coordinating” that effort? Surely everyone except the datacenter builders and the big hosted AI models has exactly the opposite incentive.

That’s exactly the point.

Wrangler, being cloudflare’s primary CLI tool, is a microcosm of exactly the problem GP was articulating: it’s focused way more on adding new commands than improving existing ones.

Many products, even supposedly “GA” ones, still lack basic operability via wrangler because instead of finishing building out its capabilities to manage existing services, they prioritized adding rudimentary support for new ones.

Now you are retreating into semantics :)

My original statement was that life expectancy in many countries is in the mid-80s. Which is, in fact, true.

If you want to split hairs about what constitutes “mid”, you can, but I doubt many people would agree with you that age 84 does not qualify as “mid-eighties”.

Do you have any evidence that the code quality of OpenClaw is to blame for its decline in popularity?

I would say far more likely is that its creator was acqui-hired and Anthropic banned OpenClaw usage.

The reality is that AI is both capable of producing sloppy code and capable of cleaning it up, if directed to do so, just like humans.

And, just like humans, code quality is very rarely the make or break factor between success and failure in business, much less popularity.

You say my claim is not true then proceed to acknowledge it is true. To be clear about the fact: about 1/4 of all countries have a life expectancy >80, and as you said, more than a dozen in mid-80s.

Life expectancy in many countries is mid 80s.

I don’t think that’s an accurate or useful characterization of modern AI like Claude at all. It is not simply regurgitating knowledge. It applies its knowledge to create bespoke solutions to the problem you pose to it, and is able to self evaluate its progress towards the completion criteria. If you don’t think that counts as “problem solving”, your definition would exclude nearly all knowledge work and engineering.

physical reality - finite supply of limited resources whose demand exceeds supply - has everything to do with the reliability of a service

The article does not claim they have achieved recursive self improvement... just that it appears to be a plausible outcome given the progress of AI development in the past few years.

I don't know about you, but AI advancements have brought extraordinary improvements to me personally in my ability to be productive, in much the same ways the article outlines. I find it deeply satisfying to be able to "get ideas out of my head" faster and tackle more meaningful problems.

FWIW, it deeply concerns me how much power and capability is being centralized in the hands of so few, especially Anthropic. I, for one, hope these advancements can be scaled down to something I can have full sovereignty over and trust... in my own home.

Presumably the bottleneck is not software correctness... even true AGI change the laws of physics (or make datacenters appear out of thin air) ...

This is the beginning of thinking, not the end...

Are you saying this result is uninteresting and therefore AI slop or puffery? Obviously OpenAI has a motivation to "market" the accomplishment as much as possible, but surely you agree it IS a remarkable achievement?