I mean it's just people saying stupid shit and scamming. You can't ride a SaaS on pay-once, ever. At some point he has to charge more or shut it down. At least the free open source guy you can self-host it, which makes some sense (though time is money, friend -- and managing that deployment isn't free) I'm so glad that like half of AI is just "what if we scammed people?"

Always wanted to throw these stats on our site, not sure if it'll be a dumb endeavor or not, but here ya go.

Number of times I've typed a prompt into the fucking terminal window right below my prompt window: 2,403,230 times.

I mean as someone that publishes on their platform, they provide a lot of tooling, exposure and functionality that many other publishers like Epic do *not*. They also provide my customers with a generally better experience, I approve of them having the ability to get a refund protected. And while I'm roll around and gripe that Steamworks *really* is in need of an update, a better UX, more metrics. What we do get for distribution management is crazy. And at the end of the day, it's my choice as a developer. I can publish across multiple platforms, I can self-publish. None of this is *harmed* by publishing on Steam. The only thing I disagree with is their pricing policy, which I'd like to see an anti-price parity law, this also has the double whammy of knee-capping Amazon. The truth though even if that got made into law? I'd just price it to cover the fees of the platform, and I 100% know a majority of our customers will still pay that to be on Steam lol

Because it was for them It still is, but was back then for them too.

I did a big project on Y2K in middle school, been a computer nerd since before I went to kindergarten. Basically summed up as "We patch our systems, anything that isn't patched has date bugs, we fix those, we got the critical stuff covered" Eyup...

Quality. We've had multiple other vendors fired and their work given to us just by not being dumbasses that have to ask AI for everything in the middle of a meeting. TBH: The gap is widening, more vendors are responding with AI slop and frustrating customers. I've had more interest in the past 6 months then probably the past like 3 years. AI is an accelerator, and the stupid come out way more stupid. I had a customer's engineer mention how he'll use AI to work a project and everyone groaned because they know the quality of his work... AI agent or not.

Like everything else we learn about agents, like with people, structuring your project so everyone isn't constantly clobbering the same files helps *a ton*.

LinkedIn has been a hole of basically almost useless bullshit to me. It's endless circlejerking, and these days mostly AI spam. Basically zero meaningful or insightful discussions. I've gotten more leads from random other sites. I mostly use it to cross-reference resumes but that's it.

This is where you make a linkedin post about coding a replacement to save $7/mo

It's a major struggle, our industry is full of dorks like this, people who will get hired by some CEO who barely knows how to do anything. On top of that, American corporate culture basically encourages you to play nice, not call out any of this, people will get angry. So you let the guy ride along, suck value out of the company, bounce when it doesn't work. We'll try again. It's fucking bonkers how often this shit happens. I've got many stories.

The number of dumbass business owners that would buy bottom-tier cheap hosting then roll around and complain that they're not getting a massive amount of effectively free support is too many. "Well maybe support can write a script..." No that's not their fucking job, then they'd act surprised that they wouldn't, then they'd threaten to leave which no one fucking cares about your $40/mo VPS. There's a flavor of business owner that is an intolerable idiot.

I enjoy that all of these “experts” are just discovering side channel attacks, or are surprised that AIs can exploit ones we’ve known about for nearly two decades. In the field this is mostly a non-issue because it’s so impractical. Good luck reliably landing on the same AWS host as your target (who is also running your software). Which is also why “sandboxing” agents on shared hardware, telling them to break out, and then being shocked they use a technique that’s all over their training data is stupid. We’ve had white papers on this FOREVER. We know it works, including in noisy environments where the hardware isn’t dedicated to the experiment (harder, but still doable). I’m so tired of my industry running on Knoll’s Law of Media Accuracy.

Claude: "We can validate by doing [x]" Me: "Isn't that going to be slow as fuck?" Claude: "No, it'll be fine" Query in production: *times out after an hour* Don't worry, programming is solved.

Big oof

"Getting nowhere" is something that's maybe a bit concerning, shouldn't porting be one of the easier things to do? At least in my experience porting gets me the furthest with the least effort of any of the projects I'm usually working on. The gap in performance is crazy too, especially with all the good things I hear about Sol and Astra. One of these days I may have enough free time to actually poke around at models and not just await getting forced off by someone making a bad pricing decision, oh well.

I just had to push Opus 5.5 multiple times on what I was suspicious was insecure upon reading the code and how it was handling things. Turns out there's two injection attacks against it. I'm not saying you have to read every line, but know when you should look.