barrkel

9 ms·

121.
by barrkel 7mo ago
There's a tautology encoded in the question. You become an adult when you behave like the people who most people consider to be adults, behave. What is an adult? Like most words, "adult" encodes a cluster of related behaviors
122.
by barrkel 7mo ago
Single biggest problem with doing compute on a laptop is it stops when you shut the lid.
123.
by barrkel 7mo ago
For batch jobs there isn't much competition. 9995wx has 3 to 4x throughput of M5 max. And then, if your laptop is busy, your machine is occupied - I hate that feeling. I never run heavy software on my laptop. My machine is in the cella
124.
by barrkel 7mo ago
Any and all. It's not particularly justifiable. It's more like, I'm a software engineer, and this is my home workshop. I run dozens of services, experiment with a bunch of different LLMs, tune my Postgres instance for good pe
125.
by barrkel 7mo ago
Sure. You're right, it is my money. And I pay even more for inference on top; I have OpenRouter credits, OpenAI subscription, Claude Max subscription. It's not so easy to get nice second-hand hardware here in Switzerland, and my H
126.
by barrkel 7mo ago
Alas, I'm not a young man any more. And my HEDT is headless, it has no monitor with which to play FPSes.
127.
by barrkel 7mo ago
Look at the way age gating is going in a global coordinated push. Can control of compute be far behind? It wasn't my primary motivator but it hasn't made me regret my decision. I hummed and hawed on it for a good few months myself
128.
by barrkel 7mo ago
People spend a lot more than that on a car they use less, especially if they're in tech. The RAM choice was because I have never regretted buying more RAM - it's practically always a better trade than a slightly faster CPU - and 9
129.
by barrkel 7mo ago
I don't buy the central thesis of the article. We won't be in a supply crunch forever. However, I do believe that we're at an inflection point where DC hardware is diverging rapidly from consumer compute. Most consumers are u
130.
by barrkel 7mo ago
I am now wholly bought into the idea of having a feature branch with (A->B->C) commits is an anti-pattern. Instead, if the feature doesn't work without the full chain of A+B+C, either the code introduced in A+B is orphaned except
131.
by barrkel 7mo ago
By "unit of work", I mean the atomic delta which can, on its own, become part of the deployable state of the software. The thing which has a Change-Id in Gerrit. The delta is the important thing. Git is deficient in this respect;
132.
by barrkel 7mo ago
I don't really get the upside of focus on CRDTs. The semantic problem with conflicts exists either way. You get a consistent outcome and a slightly better description of the conflict, but in a way that possibly interleaves changes, whi
133.
by barrkel 7mo ago
The abstract is over-egged. The language obscures what it purports to find. So some prompts return null results. "Ontologically null concepts" could just be a fancy way of saying "the model doesn't know what to do with n
134.
by barrkel 7mo ago
Christy Kinahan is still at large, in or around Dubai.
135.
by barrkel 7mo ago
A false representation of user interest
136.
by barrkel 7mo ago
This, for me, is a current design flaw of Postgres. You expect your database to trade off memory vs spilling to disk within the resources you give it and the load it's under. Databases are much like operating systems; filesystems, sche
137.
by barrkel 7mo ago
There's essential complexity and accidental complexity. A sufficiently detailed spec need only concern itself with essential complexity. Applications are chock-full of accidental complexity.
138.
by barrkel 7mo ago
I can taste the difference between a $100 wine and $400 wine, but it's maybe 20% better, if it's possible to flatten extra layers of flavour into a linear scale. It's easier to appreciate for different levels of quality from
139.
by barrkel 7mo ago
Claude munches through Ruby just fine, all day long.
140.
by barrkel 7mo ago
As an owner of a 96 core 9995wx, nobody is buying one for desktop PC much less laptop level software. To justify the investment you need to have tasks that scale out, or loads of heterogeneous tasks to support concurrently.
141.
by barrkel 7mo ago
I always tell Claude, choose your own stack but no node_modules. What's missing is another LLM dialog between you and Claude. One that figures out your priorities, your non-functional requirements, and instructs Claude appropriately. W
142.
by barrkel 7mo ago
Apart from dexterity, bipedal machines are unstable and require dynamic adjustment to stay upright, as I understand it. The mechanism humans use to stay upright after an unexpected loss of balance, flailing etc., would not be safe to be aro
143.
by barrkel 7mo ago
World War 1 was not the kind of war that delivered freedom. It was more the kind that elites entered into without full regard of the costs.
144.
by barrkel 7mo ago
This has the feel of onanism, sorry.
145.
by barrkel 7mo ago
What about the cost of k8s and AWS experts etc.?
146.
by barrkel 7mo ago
$50 will last you a good long time if you don't use many tokens, and you are judicious about which model you use, and don't need web search much. However, on a fixed price plan your behavior changes. It's a qualitative change
147.
by barrkel 7mo ago
Demand (value created) isn't enough to make prices rise. You need supply to be constrained. If there was only one competent coding model, I'd be worried. But between competition and open weight models, we're not looking const
148.
by barrkel 7mo ago
The purpose of tariffs, contra Trump warblings about raising money, is to divert expenditure from imports to domestic alternatives. They are a not very effective tool to try and fight the US trade deficit, and in particular, mercantilist po
149.
by barrkel 7mo ago
In my view, power consumption isn't relevant to a desktop or workstation (and increasingly, desktop machines are workstations since almost everyone uses laptops instead). When I'm plugged into a wall socket, I will take performanc
150.
by barrkel 8mo ago
Rtx 6000 pro Blackwell, not ada, for 96GB.

More ›