You can't, "agents" and "agentic coding" are incompatible with the deep understanding of your code that getting into the flow state requires. By definition, prompting a tool to do your job for you means giving up the depth of control that comes with actually writing the code.
Some people can work like that, some very few people may even produce acceptable results like that. I cannot and I am not interested in doing so.
The closest you can get to agentic coding while still actually being in the control loop is something like intellisense - where it just fills out the implementation of small chunks of code, one block at a time and only when you specifically request it.
> “The biggest issue is a belief that AI should be how we solve everything, while ignoring the resources that it costs. This culture is omnipresent across tech.”
Well said. It reminds me of the peak of crypto hype, but worse and more pervasive. There's this attitude that no matter what the problem is, the solution MUST be LLMs.
Datacenters (even the new ones being built today) do a lot more than just AI, though. If the goal is to push back on AI, a moratorium specifically on datacenters rated for more than X megawatts or more than Y acre-feet of daily water per acre of occupied land would've been a much more reasonable approach here than just banning datacenters entirely.
> The council approved two measures: an ordinance halting applications for data centers with electrical capacity of more than 20 megavolt-amperes — enough power for thousands of homes — and a resolution committing the city to study their impacts as a precursor to permanent regulations. [0]
Well, looks like this is exactly what you're asking for, actually! They have passed a moratorium on datacenters drawing more than 20 megavolt-ampers [1] and have committed to gathering more data before passing permanent regulation.
So, since what the council actually did is exactly what you're asking for here, you support the decision of the Seattle City Council to pass this, right?
No, because that's per datacenter, not per datacenter-acre. I'll give them partial credit for trying, but a 5-acre 20MW datacenter and a 50-acre 20MW datacenter are not going to have equivalent use cases or local impacts, and that'll make measuring those impacts more difficult.
Your desired unit of measurement is a stupid and pointless one.
No one measures data centers by “datacenter-acre”, because what matters is resource consumption not power density. A 20 MW is going to consumer 20 WM of power and require 20 MW of cooling, regardless of how many acres it sits on.
Power density is the exact metric by which datacenters are differentiated between “hyperscalar” (i.e. what most people call “AI datacenters”) v. your run-of-the-mill colocation DC.
Also: a flat MW cap per DC is straightforward to game by splitting one big DC into multiple smaller DCs.
The entire process has blind spots. Starting with adding more demand for electricity, more demand for mined materials that harm the people and environment. Amazon hasn't helped the global environmental footprint in any way. If you work in most forms of tech, you've agreed to not lose sleep over these things.
No matter what the problem is, the solution will involve some form of machine learning, inference, or both, with massively-parallel processing that is probably (and unfortunately) centralized to some extent. Hence the need for data centers.
They don't need to be built in the middle of downtown freaking Seattle, though.
> No matter what the problem is, the solution will involve some form of machine learning, inference, or both, with massively-parallel processing that is probably (and unfortunately) centralized to some extent.
I think the point people are making is that this claim is not self-evident, and there's a remarkable lack of justification for it whenever it gets asserted. If you're convinced that this tool can literally solve every problem we have in society, it would help to explain why you're so confident about that. So far all I've heard ever is "exponential growth", which is not particularly convincing when a high school precalculus class gives you enough knowledge to be able to understand that there are curves that look a lot like exponentials before suddenly hitting diminishing returns.
Google doesn't care about reputation the way Apple does. People are used to Google shipping broken features and shutting down products when they don't immediately sell. Apple is a premium brand and it's expected to be more stable, and to deliver solutions that actually work.
Well they're paying Google for access to their model, which I think makes sense. I've been saying for a while that models are commodities, the integration or harness is what really matters. There's a lot of good models available, many open weight these days!
What took Anthropic to the top wasn't their models, which are good. It's Claude Code, everything around the model that delivers value for their customers. Gone are the days of slapping a model behind a chatbox and hoping for the best. We need, like, actual software, that does actual things. Siri could be that... maybe. Apple has a lot going for them in regards to how much of their consumer base's software experience they control.
Genuinely, the day they start shoving "AI features" into my iPhone is the day I switch to a fairphone or something I can actually own. I don't care if they include optional features, but I am so tired of seeing chatbots and text manglers shoved into every fucking app. I can write my own emails, thanks google.
You have anything to back up that claim, or is it just knee-jerk drivel with no evidence based on your feelings and distrust of scary government bureaucrats?
And how is that different from any other governmental level? Seriously, we in Berlin got a freeway that nobody I know wanted or wants. It's not the EU's fault.
I think some do use regular wifi, especially the amazon ones, but with a button you have an advantage where you could just stay idle and off the network most of the time, and only connect to send an event when pressed.
That likely increases your latency by a fair bit. Most of my devices take a second-ish to connect, configure the WiFi stack and start sending data.
I’m not sure I’d want a full second of latency on button presses. I would have figured Amazon used BTLE to the Alexa and used it as a BT to WiFi gateway
BTLE and/or their IoT mesh network (Sidewalk?) would be a good way to go too, but if it's just a button to reorder laundry detergent or something then I don't think even a 30 second delay would be a deal killer for most people.
In 6 months they'll be in the comments on the next article about covid railing against the vaccine again. It's never enough with these guys, because it's not about being right or wrong, it's about having their feelings validated. They feel like the vaccine was bad, facts be damned.
Some people can work like that, some very few people may even produce acceptable results like that. I cannot and I am not interested in doing so.
The closest you can get to agentic coding while still actually being in the control loop is something like intellisense - where it just fills out the implementation of small chunks of code, one block at a time and only when you specifically request it.
reply