petesergeant

I wrote a very similar article called "The prompt is not the work", which also touches on honesty about creation, and disclaimers on where your contribution lies: https://sgnt.ai/p/prompt-is-not-the-work/

I think we have a couple of years of "being good at talking to the LLM about your field of expertise" being a useful human skill, until that too gets washed away

At least on Windows, the most you get is a scare screen

Well that and side-loaded LG adware

One can't expect that every driver package submission from every vendor is thoroughly vetted every time, this is a matter of trust and respect.

Apple seem to do a pretty decent job of just that

Who is responsible for the crimes of a "rogue" agent? How will they be punished?

Unironically this is why AI researchers have this fascination with the Talmud.

I believe this is called the “circular economy”

Can you explain why you think they would?

and is now doing rather well

I was going to write a shitty reply to this, but the more research I did, the more it seems actually this went pretty well for them. UK and NL governments that lost deposits in the Icelandic banks mostly got their cash back (eventually), there was recession and unemployment but not that much worse than other countries, and the country is in good standing again with the markets.

I would note that this is much much easier to do if your investors are foreign, rather than domestic pension funds, so it doesn’t bring down the government, though.

In theory Cerebras have a developer subscription model you can use, but they seem to have stopped new signups. So per sibling, OpenRouter and per-call pricing is the answer for now.

LLMs are inherently adversarial (read, "relentlessly proactive") [and] you still tend to end up with something you care about on the same side of the trust boundary as the LLM

I mean this is a problem with many coworkers too, so you deal with it in the same way: limit what they can do to creating pull requests.

Thank you!

You should absolutely be running your AI agent inside _some_ kind of sandbox. I put together a list of 19 mostly open-source ones here: https://pleasedonotescape.com/ along with a list of non-project-solutions

flagged per the highly misleading title

This is not how it’s going right now, and FWIW I also don’t think it’s going to go that way, but it’s certainly a plausible investment thesis.

The valuation is based on one lab getting a decisive first advantage, and turning that into a durable self-improving advantage that can never be caught up to. If any can pull it off (a gigantic if), they will effectively own most AI value, and the people who own their shares will live happily ever after. Divide your investment between the labs that could plausibly do this, and your EV may not be dreadful.

Habibi, I don't think anyone is really blaming you personally. We are piling on to the fact that this piece of critical infrastructure many of us depend on day in day out is being built in such a way that a single developer can wake up one morning, ship a change they thought might be nice, and then walk it back the next day.

Yah. And it's not like they can't afford the talent to do this right either. I've said elsewhere, I think it's an attribution error. Claude Code is massively popular, but arguably because of the model/subscription, but I think the brass reads this as "great success throughout"

Also: https://news.ycombinator.com/item?id=48766895

I would love Claude Code to be a little less vibe-coded. The underlying model is excellent, but we're being pretty much forced into using CC to use the subscription model.

https://github.com/pjlsergeant/byre can add small but very convenient layer around whichever agent you want, initially trapping it in a specific folder, but you can easily mount more, firewall it, bring in MCPs and skills, or whatever. It's unambiguously AI-assisted, but a very very great deal of thought and work has gone in to the design of it. Ask your favourite agent to review the code if you're skeptical.

closing at $135.27

ffs, wake me up when it's at least 10% below what it IPO'd for. The idiotic tulip mania that followed in the few days after it floated was noise, but as of today, it seems the IPO price was pretty much right. However, endless headlines about the price crashing etc.

From a fundamentals perspective, it's an insane price, obviously. But the narrative that it's all coming crashing down is obviously not correct (today).

almost definitely requires an LLM to tackle at all

Conveniently I have some of those… first day of trying to script Grok Build I think I sent in 6 bugs of slightly weird behaviour I discovered, it will be much more useful to (have an agent) check the source and see if stuff looks deliberate or like a bug, etc

Genuinely curious about whether comments like this consider all AI generated codebases to be slop? Are you just knee-jerking or is this one an example of actual trash? I have been building a product[0] where I’ve not written a single line of code; is it also definitely “just tokenmaxxed slop” or is any consideration going into comments like this?

0: https://github.com/pjlsergeant/byre

This feels very “I could build Uber.app in a weekend”

I have found Grok Build to be decent, and the harness to be competitive with similar harnesses. What will Cursor add if I check it out?

I’m an anti-Musk zealot and I now pay for a subscription, it’s pretty good stuff

Neat, trying to reverse engineer some specifics of how it does stuff has been a pain in the ass, and this will make it easier.

It's hard to vibe code security.

Conversely, running human or LLM-generated code through multiple LLMs to look for security holes is a fantastic way to increase security.

Classification of types of user frustrations and sentiment analysis, content trends, engagement and gap analysis, as well as then looking at changes from the previous week. We also look at how certain queries turn into actions in the system (eg: which users take actions we offer them). We run it once a week, rather than every day, and it provides an exec-facing overview, as well as areas for support to dig further in to. While it's some good work, as far as I'm aware it's almost all just a text prompt and a connection into Langfuse.

My junior developer has a Claude Cowork skill she built to do this over about 25,000 messages a week to our agent, and it seems to work pretty well. Struggling to understand what $499/month would buy us here?

In my experience they're pretty well aware of this failure mode, to the point where I learned the word accretive from adversarial agent code reviews this week. Less good at fixing it, but once you know it'll happen you can ask it not to.