This is bad advice in 2026 for most people who would read it, since it advises taking a terrible security posture (give the agent access to everything,) in exchange for a relatively small improvement in workflows.
I say small improvement because my experience is that modern Agents are pretty good, so by the time they've handed it back to me to test it, there are usually only one or two remaining issues that I'll discover as we roll it out to Production.
(In general, I think we don't do enough to emphasize best practices in the era of AI, but...)
What Kent completely ignores here, as far a I can tell, is that there is significant value in finding out sooner what the needed features are. Building speculative structure can be a forcing function to establish requirements, because at least you start exposing failure modes. It might be more expensive than waiting, so hopefully you don't do it for most of your requirements, but sometimes it's your best option.
Building the wrong thing is now a much less expensive option, and that means the calculation around YAGNI is different. But it's still a calculation, and for now, each team needs to figure out how it has changed for them.
This is my concern as well. If critical open source packages become dependent on these corporations for "secure" releases, does that enable them to force ID verification into packages, for example?
Related, but most of the smart folk I know think Open Source AI means Anthropic and OpenAI are financially impossible. A lot of the companies signed onto this are heavily, heavily leveraged by those two, and have significant incentive to disrupt Open Source AI before all their customers get sticker shock. I've been waiting to see what their move would be, and this might be part of it.
They default it to talking to a free version of their model (which is incredibly cheap if you decide you like it.)
But it seems trivially easy to run it against local models. Their onboarding guide offers that option, though I have no idea if it changes any functionality.
Interesting. I've been using their models quite a bit, through Claude Code, might try switching to this.
I also built a couple of harnesses, I wonder if I could swap this for Claude in those...
(Lots of interesting ideas in the blog post, but like all the AI developments these days, hard to be sure what's valuable without extensive experimentation.)
Because this isn't the first collection they've stolen from someone, presumably.
It's a lot easier to become a really successful company if you can keep your inventory costs down. Perhaps by investing in local law enforcement instead, to make sure no one looks too close at said inventory?
Donald Trump is famous for not paying even really cheap contractor bills, because he knew he could get away with it.
People, especially in remote jobs, benefit from being organized into groups intentionally, with distinct rituals that enable them to operate effectively while they get to know each other better. Another person needs to design and oversee all that.
While you can provide templates for that structure that allow oversight to scale so that one person can oversee larger groups, that tends to be more effective in non-remote, and more predictable, work environments. Modern software development is very little of that.
I don't have much in-person experience with middle management in contexts outside of software development, and I suspect there are some opportunities to use AI to bring engineers closer to customers.
For whatever reason, I get very few push notifications on my phone. Compared to my days at Blackberry, it's probably 10% as frequent that I get interrupted by my phone.
So good for me.
But there's some really scary stuff in here happening to other people that I'm not even aware of.
That's not generally true in the US over the last 40 years, where the gains from productivity increases have been accumulated almost entirely by the top classes.
Yes, lower classes have access to many more conveniences then they might have had in earlier decades, but they are working far more hours, and their expected lifespan has started decreasing.
I was using Gemini and Claude quite a bit over the last year, mostly with pro or opus for planning and flash or sonnet for implementation.
MiMo is the best one I've used so far, but I haven't done anything interesting with the Claude 4.7 models.
It seems conservative with generally good "instincts", getting things working quickly without too much complexity.
I've also embedded it in several different projects so far, and it's been pretty easy and effective.
I think this does a very good job of describing the real gaps agents are hitting in practical usage, along with a fairly compelling rationale for why those gaps aren't likely to disappear any time soon.
If we're going to stabilize the software industry, we need to have more discussions like this that identify what constraints apply. (We should have had those discussion before pushing AI out this widely, but that wouldn't have gotten anyone rich.)
I actually think that there's a world of software systems agents can change, but it's materially different from the one we have now, and has a different set of constraints that we've also mostly done a poor job identifying. So hopefully the discussion can help those of us on both sides. ;)
Oh, I don't want mean for myself, I mean as an experiment at the other end of the four-day work week spectrum.
Though for myself, I like having time to think about ideas in between when I collaborate with folk on them, so I have a lot of optimism about the success of the experiment. :)
Take all that AI productivity and found a one-day work week company. One day of focused collaboration each week, let bots and brains chew on stuff in the interim.
I've gone through ~600m tokens in Xiaomi Mimo though Claude, and it's been the most effective use of an agent I've had yet. It's very capable, but generally not ambitious, picking simple but effective solutions to most problems I give it.
Going to write something longer about the experience when I get to a billion tokens.
I say small improvement because my experience is that modern Agents are pretty good, so by the time they've handed it back to me to test it, there are usually only one or two remaining issues that I'll discover as we roll it out to Production.