He also died crossing the tram track, presumable not looking both ways before crossing. To be clear, I've also nearly been hit by the Barcelona tram too, so I don't blame him, but "smart" is always relative.
It's specifically for "jobs in sectors like agriculture and construction", not sure better buildings will help the workers out on fields and construction sites much.
Although one temperature threshold probably don't make much sense, a strict 30C across the entire continent would mean massively different things in different countries, in many countries that mean basically the entire summer off, every year.
Never had that happen in ChatGPT itself, I almost always use Pro mode whenever I use ChatGPT, but what you say happens a ton in codex, when I look through the session traces it seems to happen because of the automatic compaction, where some assumption the initial pass did gets passed on as a question from the user to the part after compaction, which is a bit confusing. I think it was mentioned somewhere that the compaction got a lot better, but I haven't used GPT-5.6 enough to say if it's actually better or not on that.
This post feels like how my own notes are done before they're polished into a proper "article/blog post" or whatever, and I genuinely appreciate what seems like a stream of consciousness from the author made public (not sure I'd dare to), even though I don't personally agree with it all.
> The new war demands your inner reality. The new war will be weird in all sorts of new ways we can’t even imagine yet.
I've been orienting myself towards this already being true as well, and think we still haven't even started to see this taken to its (logical) extreme. If nothing else, it'll at least be interesting to see all the effects and methods around this, and all the cool mind reading toys.
I've found this to be true too, that's why I've moved most of my personal email to my own email servers, so at least I can decide when those outages happen, as they tend to mostly happen when you do some changes, not just randomly by themselves. At once you've setup monitoring and recovery for the usual suspects (disks, network, etc).
> That sounds like a good idea, but shipping 10x as many features and bugfixes sounds better.
This work great until you reach a certain size, then good (or even "not bad") code is required otherwise the model spins its wheel trying to ensure the change is correct.
The way I've measured how good/bad the code is (for AI) is to have one "baseline fixed change" that I measure how long time it takes to implement. Always in the beginning (less than 10K LOC, as just some measurement), this baseline change will take 2-3 minutes. As you add more code, the same change starts to take 5-6 minutes, and once you hit 1 million LOC, it can take as long as 10 minutes, even though the change is the same.
It's when this baseline task starts to take longer time, that you need to update the design/architecture/layout/whatever, to better fit the task/domain, and to actually make it easy to maintain and still possible to add changes without spending 10 minutes. So its at this point you refactor, and once done, the baseline task will again be easy for the model to do.
So yeah, if all you do is smaller projects, then "shipping 10x as many features" is easy and doable, for the lifetime of the projects. But once the projects start to accumulate technical debt, the model will have a harder time making sure the changes are correct, and suddenly "shipping 2x as many features" is maybe doable, but you could still have had 10x if you just spend slightly more time on the actual design and architecture of the program.
> You do that because your goal s “optimized code” not just “code”.
Again, my goal is "Ship working software", I don't care about optimizing at all, just baselevel "working" state of things, that's why I test things like reliability. But also again, our perspectives and understanding of terms seems to diverge a bunch, so much that discussing them seems relatively fruitless, at least seemingly to the two of us.
> Many people run local LLMs which is where a mac is useful
Unless you go for the very expensive options, most of the Mac Minis really aren't suitable for running local LLMs, they're painfully slow with prefill/processing input, and the models you are able to run don't handle long context very well, which these sort of long-running agents perform very differently with when you can.
I'll agree with your latter point, hard to beat the value of using something like OpenRouter or similar remote inference.
Even with local models, you can run the agent software and the inference workload on different hosts, which is what I'm doing at home. Beefy server responsible for inference, tiny VM on other server is running the actual agent software + RPC + bridges and what not.
> "people often want a system that's under their control, isolated from their primary machine, and capable of running 24 hours a day, seven days a week," said Brooks. "A Mac mini is an amazing system for that," he added.
These execs are so out of touch they believe Apple hardware to be "a system that's under their control", how does it come to this? Besides, a VM without bi-directional sharing of data gives you pretty much the exact same thing.
Did hundreds/thousands of developers really go out there and bought Mac Minis just because one prominent technology semi-celebrity happens to have used a Mac Mini for the development of their thing? Seems bananas people would spend hundreds on monies on something they barely grasp how it works.
> I didn't expect this simple concept will need so many explanations.
You and me both brother :)
> No programmer will ever say "making the computer slower really helped a lot to write code".
I guess I'll be the first, as a programmer I do this all the time, when developing browser stuff I frequently throttle the available bandwidth and introduce jitter in the network connections so I can see and understand how things work with less optimal network connections, loads of us professional programmers do this when aiming for high quality software meant for end-users, in loads of different environments, not just browser development. Making the computer/IO/whatever slower/"less" does help a lot to write software for others.
> You're expecting programmers to introduce things that prevent them from reaching their goal of creating programs that run.
Well, yes and no. I expect professional programmers, my peers, to do this, as this is how you typically build reliable software, but I don't expect everyone to do this, definitively not amateurs just programming for fun.
> Your goal is "make a creative song" but you're stuck. You don't introduce constraints because they'll make it harder to get unstuck, you introduce them to make it easier. You literally said "they still help a lot".
Cheers for the attempt to decide what my goal is but no, the goal never is "make a creative song", the goal is precisely as previously stated: "make a song". Helpful tip for the future, read and understand what people write and don't assume they lie or misunderstand their own intention, and it'll be easier to understand other's point of view.
Regardless, I feel like both of us are digged into our understanding what the whole "add constraints to do better" process means and is, and no harm no foul, what I know helps me and I'm sure what you know helps you, so lets just leave it at that :) Wish you the best of weekends!
Yeah, all the labs seems to converging into the same (post)training for all models, while in reality, different user groups have wildly different requirements and expectations from these models.
I want the same as you, and even further, I want a model that refuses to execute changes I request if they don't make sense considering the context, or if they're impossible, and avoid any sort of quick hacks and patches. But I also want a model that does the pure opposite, that I can chuck a "Do X" query at and it figures it out. Then I'm sure there are middle-zones between these two, or even more extremes too.
But the choice isn't there, we get to chose between "fast/stupid", "medium/medium" and "slow/smart", then that's it. With system prompts we get to steer it a bit, but I've needed to make my own fork of codex to surface those things to me (the user) so I can control it better, and different models respond differently to the "Stop and don't implement anything if the request doesn't make sense yadda yadda" parts, would be lovely to have those sort of "personalities" surfaced up front when making decisions about what model to use.
> So you're not introducing constraints, you're creatively trying out things to fix your problem. They're not a wall preventing you from reaching your goal, they're the bridge. Your constraint is the temporary lack of creativity, and what you introduce is the creative solution.
I feel like we're talking past each other. Adding these sort of requirements in order to "fix the problem", is typically what people are referring to as "adding artificial constraints to foster creativity".
The goal is making a song, anything that restricts you on how you are allowed to do this, is a constraint, as far as I understand the word "constraint" at least.
> When you remove actual constraints you make the goal easier to reach.
Yes, this is why the previous examples are constraints, not "bridges". Without them it's easier, with them it's harder.
> For example, it's trivially easier today to create a 3d animation than it was 5, 10, or 20 years ago, and the visual quality of it would be similarly disproportionately better.
True, but this also means that the bar has risen for animations in general, so while you might be able to create animations today as an amateur that is even better than the animations just five years ago, it still won't come close to what professionals can actually achieve today.
Your very last point speaks a lot to me though, almost every effort people are amazed by have at least two people involved, indirectly or directly, and attempting things like this on your own would be a fool's errand.
> Constraints you can remove with a flick of a finger are not constraints.
Why not? Who cares how/why they're there, as long as you follow the constraints, regardless of how easy they are to remove, they're still there.
I frequently use this when stuck creatively in music production. "Ok, now I can only use this filter for any sound shaping", or "Make a song using only instruments outputting mono", or "Maximum 10 cables to make a new sound on the modular synth" or whatever. Really easy for me to skip these artificial constraints at any time, they still help a lot.
Steam (which I'm guessing you're talking about) is nowhere close of being a monopoly. There are loads of alternatives out there, in wide use by people already. World of Warcraft, League of Legends, Minecraft, Roblox and more are all examples of big time successful games that never been available on Steam.
Indeed! And I bet Apple technicians have even better training, tools and reference documentation, despite that they broke my screen anyways! So I'll stick to my warning to ask people to be careful, seemingly really easy to break the displays in the process.
I think the article kind of gives you the energy you enter the reading with.
My take away was closer to "This is what we tried, this is the sentiment they gave back, then this is what happened, here's what I think of Jarred's way of working, glad they now use Rust, good riddance", but then I also apply this slightly unholy strategy of applying charitable reading when reading personal blog posts, and I worship neither of these people, so might be why we got slightly different takeaways.
https://emsh.cat
[email protected]