poisonborz

Thanks for your service! A fair warning: I would advise to move off of Proton due to recent news / CEO behaviour.

I don't get this comment, you say it breaks trust because it's unknown who is accountable, and then you say you (the author) is accountable. So where is the trust broken?

Also your examples are wrong, regardless of pair programming or TDD, there is always a single committer, who is to be held accountable. How you achieve that commit in the past, copy-pasting from stackoverflow, or some auto-expand snippet automation from the IDE - so any tooling - does not matter. Agentic development is also such a tool, thus does not change trust.

The agent runner, or committer, a single human, is to be held accountable. That responsibility is - and always was - the most important role of a programmer.

- large enterprises, especially if their main business not software, are rather cagey about new contracts, and it can take 1Y+ to get a contract approved

- I'd never blow this budget to be dependent on a single provider that 1. can change contract prices or limits on a whim, 2. lives in a sea of competition, new models coming up every week that may be better for a certain workload.

Claude lovers tend to be either early adopters not well informed/lazy to experiment ("why, it's still the best, right?"), or millionaires or devs with unlimited budgets, who often ask on HN "why would you use anything else than Opus on xhigh".

I was handed 150$/mo budget and no subs, and while it was frustrating at first, with better tooling, context management and especially OAI models it's a breeze now and not even using 70% mostly while achieving a lot in a small team.

Having limited resources teaches you a lot.

I think I reacted exactly to this. What is "code as craft", or as often read, "code is poetry"? It is variable naming, structure, hierarchy, import logic, all patterns that can be strictly controlled still. Opinions vary if this still carries enough value - I like these details, exactly because I value the "craft" and the expectancy of human oversight these presume, but I could argue about it to some manager less and less (compared to what frontier models generate already).

So when all this can still be kept, "hand craft", the muscle memory of typing code, or, as OP said,"dev who liked more the micro parts", what is this other than some nice myth the industry wove for itself?

Apples and oranges. Hot take: fine details in an artwork are extremely important. "Fine details" in a codebase are either some pattern you can easily instruct an agent to follow, or deliver zero value.

Since you seem to be a high profile user, I wonder, do you not have a fear of basically giving away the keys to your kingdom to a corporate entity? That first, second, third parties or malicious actors in or outside the supplier chain might use this information against you, or the likes of you? Or simply, what kind of society you support by using and advocating for this service?

Both tasks seem to be like a job that in the older times could have been handled by a personal assistant / a human agent, probably way more expensive, but the threat vector seems to be so much simpler.

It amazes me how even smart, capable, well paid IT people give away their most private data "because it's cheaper".

There are talented and objectively nice software engineers working for Meta, Palantir, Flock, you name it. I knew some of them, warning about moving to the US at the early 2020s. They did not understand, just oblivious, happy about the comp, and that they could fulfill their "america" childhood dreams, mostly gathered from movies they watched on the 6th floor of scrappy socialist block building. They tend to complain about workplace tensions, high pressure... but evil things? No that's someone elses doing.

Was the art the act of typing out code? Everything else, the reasoning, remains in any depth you want. Since the same reasoning produces the sameish code I argue this is not the same as who holds the brush on the canvas.

Tailscale IS Wireguard. With all the reasonable management added. You can selfhost the control part via Headscale. Plain Wireguard can be a hassle for most - that bore Netbird, WG-Easy, Netmaker... out of those Tailscale got it the most right.

They are absolutely great except: - you need a beefier GPU for transcoding. It heats up fast under heavy load and most builds are fanless. - you start to have NAS requirements / +2 hard drives, RAID etc. There are a few chinese motherboards/minipc options now that do 6 SATAs or 4 NVMEs (via PCIE 3.0 x1) but you will be quickly constrained.

It is the perfect gateway drug setup though.

What I like is how most such vibed tools - self-aware how ubiquitous they are - don't try to capture audience or offer paid services. Everything is suddenly available without logins, share via a simple url, self-hosting is expected. While AI diminishes the power, influence and recognition of creators, I find it overall a great democratizing force in software.

Which of the rules specifically do you find silly?

Ok it'll be me to say this here: the core idea is great, but the "woman can't drive" trope is really off-putting, even if that wasn't the intention. The characters could have been anyone in any setting, yet they chose a young woman and an older brother teaching her.

> The joy of programming is dying. The joy of delivering results lives on.

This sums it up well and will ring true no matter which side of the argument you're on.

No large companies I know allows chinese models, and OpenRouter is abysmal for enterprise usage. Don't confuse personal/yolo startup usage with corporate use.

I expected a lot more such apps to appear, but it's sadly also unusable in a work environment. Eg. you can't get GH Copilot usage via API per user for organizations without admin/billing manager permissions. Also you need the amount of actual workdays left, without your holidays and vacations.

I made an own TUI dashboard that uses copilot cli (reads the output of /usage command) and gets workdays from the employee intranet site.

I wouldn't think informed people are afraid of an "evil AI", it's about then complying to prompts in ways the prompter wouldn't expect - because of weak target systens, or the agent was not sandboxed enough, or sandboxing was done carelessly. All current hacking cases are like this, imagine the same with weapon systems.

Ah, the bi-weekly "I quit Face Eating Leopard Corporation" post ("btw great people work there, they do great stuff, also my options have vested")

Even more, chinese sites are chuck full of such players for a fraction of the price in all dimensions, on/offline, tactile or touch, with quality DACs.

My life quality became so much better when I discovered loose / yoga pants. Your sizes change week to week a bit in a way that it counts, and you get into random body positions in which things get tight without you immediately noticing. No clothing should have buttons on it.

A suit is the antithesis of good clothing. Its only appeal is the socially constructed notion that it looks "good".

How is that possible? In my mind in an ideal piece of clothing you can equally do (light) physical activity, or just lay around, and is low maintenance, easy to wash, doesn't wrinkle. A suite the antithesis of this.

As many have surmised, this may have been a panic rename of Astra 6.1

Buy used! They were/are really popular so there's always a number of items selling for reasonable prices, and they don't really have perishable properties (besides power on time which is _long_).

> 33 light spots, most GU10

That's a sad situation, but even if a bit ugly - a GU10->E27 adapter cotst ~$1.

Assuming we don't speak about some vibed "one-shot-deploy" situation (for which there were options in the old world as well, with the same bad results), in a well disciplined team, why would AI make this worse?

You can still write anything by doing it either small steps, or at once followed by a lot of refactors while skimming over code and asking tons of questions / making refinements via prompts, guidelines, test guardrails. A team can still reason and whitepaper about the same things. Devs who were previously shy to ask some specific details (to not seem dumb) can now confidently survey big codebases and get insights in whatever style they can swallow.

The bigger problem I see is that all this requires a lot of communication, and most importantly writing skills, something that disappears in thin air in the last decades.

> Smart light bulbs are another minefield

Just buy Hue, problem solved - although for best results it needs custom setup which is a higher initial time investment, buth then it's set and forget.

I had great discussions in PRs, debates we would never have in real life conversations. In these times review is maybe the last holdout to think teams have control over the code. AI made also this nicer by making summaries and answering the stupid questions without shame before turning to the author with confident criticism.

But you can still do that, much easier, with plain human words.

/yes I know a coding language is 100% better at expressing algorithmical logic, but it is definitely harder for humans to comprehend and express cleanly at a consistent rate. machines do that way better/