It's the other way around, unfortunately. The senior engineers will still be useful for architecture and infrastructure considerations, as well as guiding the agents. It's the junior engineers that get nailed, because there's little incentive to hire one when a LLM does a better job immediately and costs less.
That's true now. But in the world of this article, it's also the senior engineers that get nailed. In the world of this article, all code is like what machine code or bytecode is now - it's designed to be used by the machine, not the human, because the expectation is that humans will rarely, if ever, touch it.
GitHub Copilot was the only one with absolutely insane subsidies, where they metered by 'request' instead of tokens. A request that costs 3 cents could end up burning $20 worth of tokens or more. That ends this month.
I was actually quite worried, because I've been using GHCP for large chunks of work, but the billing estimator they released shows I was only at about $150-200 a month in API priced tokens. Sure, that's a subsidy for my $20 subscription, but not insane.
Heavy use of agentic coding tools, in a responsible manner, probably lands somewhere around that $200/m mark at API pricing. Assuming that makes the provider money, I don't see that being hard to swallow for businesses employing developers in Western countries, given the hours it can save.
The real risk here is to personal project vibe coders. Building a huge app by abusing subsidized plans is ending.
If their pricing turns out to be what they claim, and copilot cli has accurate token counts, they had the best deal around.
Just today, when I wasn't being especially chatty with GHCP, I used about 12 requests to get a few thousand line changes in 3 projects I'm juggling. The last project repo of copilot I closed, in 3 hours burned 38M input tokens, 28M cache, and like 400K out. For GPT5.4, high. That's like $135, in half the day, 1 of 3 instances. No crazy tool use, just lots of docs and unorganized code. GHCP charged like 70 cents for that on the old plan.
I'm more optimistic about LLMs tracking down and fixing issues in software, even without SO/forum posts, at least for OSS. I've seen enough unique insights from agents on tricky problems to know it wasn't extrapolating from a helpful comment somewhere.
It hit me that as it's deciphering some verbose log file, it has also read through all the source code that wrote that log, and likely all of the discussions/commits that went into building that (broken) feature.
Have you tried the Import plugin in Obsidian? I used it the other day, and it seemed to do a good job with OneNote, though my ON notebooks are just simple text and images. I believe it automates an export from ON to HTML and then converts that to MD.
Once they're in MD format, you can always get your favorite agent to do more formatting and organization, which was one of the big reasons I moved away from ON.
Nonsense. WinGet has the ability to add repositories, just like any other package manager. If you want the 'approved' packages for the distro, that would be the msstore repository. If you want to use the 'community feed', which WinGet warns you about the first time you use it, it's less vetted, but still goes through Defender scans and community moderators.
If you go adding any old repo to APT, you have the same risk. You should look at how much code review goes into packages for major distros like Debian, hint, not much, especially once the initial package was accepted.
GHCP at least is transparent about the pricing: hit enter on a prompt= one request. CC/Codex use some opaque quota scheme, where you never really know if a request will be 1,2,10% of your hourly max, let alone weekly max.
I've never seen much difference with context ostensibly being shorter in GHCP, all of the models (in any provider) lose the thread well before their window is full, and it seems that aggressive autocompaction is a pretty standard way to help with that, and CC/Codex do it frequently.
>I've never seen much difference with context ostensibly being shorter in GHCP, all of the models (in any provider) lose the thread well before their window is full, and it seems that aggressive autocompaction is a pretty standard way to help with that, and CC/Codex do it frequently.
Then we've had wildly different results. Running CC and GH copilot with Opus 4.6 on same task and the results out of CC were just better, likewise for Codex and GPT 5.4. I have to assume it's the aggressive context compaction/limited context loading because tracking what copilot does it seems to read way less context and then misses out on stuff other agents pick up automatically.
Windows has great support, surprisingly, for TPM-backed sk keys using Windows Hello and OpenSSH. Protected with physical presence and anti-hammering at the hardware level, and easy to setup by just selecting a sk type key.
I only use password keys for things that need to be scripted.
You can just replace dropbear with openssh on OpenWRT. That was one of the first things I did, since DropBear also doesn't support hardware backed (sk) keys. Just move it to 2222 and disable the service.
I reenabled DB on that alt port when I did the recent major update, just in case, but it wasn't necessary. After the upgrade, OpenSSH was alive and ready.
GHCP also has magical rate limits that hit users that slam multi-agent workflows or other crazy request burners.
Mind you, I think GHCP is a great service at an excellent price, but the hardcore vibe coders complain about the rate limits that I've never personally experienced using the CLI.
That's weird, because every time I see someone even talking positively about Claude Code they always seem to mention they're hitting their 5 hour limits in 2-3 hours all the time, they're hitting their overall limits all the time, and so on.
Meanwhile I can't even seem to spend my $20 Cursor Composer 2 tokens using their agent. I've been doing useless shit just to see how much usage I can cram in there and it'd probably take 10 hours of vibecoding like a loser every day to hit the limits at this point.
With that said I'm not going to pay for something that doesn't allow me to use whatever I want to use (in terms of harness, etc.), so both Anthropic (who were already disqualified because of their ridiculous limits) and Cursor is out (AFAIK you can't an agent other than their `agent` binary without some ridiculous hack like proxying all of the calls through `agent`.
I can't imagine all of the providers pretending their agents are real value going forward, but even if they do there's still stuff like OpenRouter which doesn't give a shit, may as well use something like that.
reply