Working on a Vercel-like cloud hosting PaaS, but specifically tailored to South Africa. Everything is hosted on local servers, and pricing is ZAR instead of USD. It's called Zanode if you want to check it out :)
Yes, I also see that (also using dark mode on Chrome without Dark Reader extension). I sometimes use the Dark Reader Chrome extension, which usually breaks sites' colours, but this time it actually fixes the site.
"""
- Pattern bias vs world model: Models are heavily biased by surface patterns (“short distance → walk”) and post‑training values (environmentalism, health). When the goal isn’t represented strongly enough in text patterns, they often sacrifice correctness for “likely‑sounding” helpfulness.
- Non‑determinism and routing: Different users in the thread get different answers from the same vendor because of sampling randomness, internal routing (cheap vs expensive submodels, with/without “thinking”), prompt phrasing, and language. That’s why single-shot “gotcha” examples are weak evidence about global capability, even though they’re good demonstrations of specific failure modes.
- Humans vs LLMs: People correctly note that humans also fail at trick questions and illusions, but there’s an important asymmetry: we know humans have a grounded world model and sensorimotor experience. With LLMs, we only have behavior. Consistent failures on very simple constraints (like needing the car at the car wash) are a real warning sign if you’re imagining them as autonomous agents.
- Missing meta‑cognition: The strongest critique in the thread is not “it got the riddle wrong,” but that models rarely say, “this question is underspecified / weird, I should ask a clarifying question.” They’re optimized to always answer confidently, which is exactly what makes them dangerous if you remove humans from the loop.
- Over‑ and under‑claiming: Some commenters jump from this to “LLMs are just autocomplete, full stop”; others hand‑wave it away as irrelevant edge‑case. Both are overstated. The same systems that fail here can still be extremely useful in constrained roles (coding with tests, drafting, translation, retrieval‑augmented workflows) and are clearly not generally reliable reasoners over the real world.
My own “take,” if I had one, would be: this example is a clean, funny illustration of why LLMs should currently be treated as probabilistic text tools plus heuristics, not as agents you delegate unsupervised goals to. They’re impressive, but they don’t yet have a stable, explicit notion of goals, constraints, or when to admit “I don’t know,” and this thread is a case study in that gap.
"""
You would be surprised, however, at how much detail humans also need to understand each other. We often want AI to just "understand" us in ways many people may not initially have understood us without extra communication.
I've been using .xyz domains for a while now because of how cheap they are (around $2 on Spaceship for first year), especially as a solo dev building all kinds of side projects that would benefit from having their own domains.
I recently launched an app on a .xyz domain that's been getting steady traffic and some people actually signing up for it (it's tinytune.xyz if you're curious to see).
Doing a quick Google search I found a few other popular services using .xyz domains:
1. MEE6 (the popular Discord bot): mee6.xyz
2. Together AI (used to be on .xyz before going over to .ai, an example perhaps of starting with a cheaper domain before going more expensive): together.xyz
Thank you so much! And thank you for your question.
Yes, so to answer it, the idea of TinyTune was to literally be "tiny", i.e., very simple to use. It (mostly) takes 3 steps + some waiting time and you have a custom model. I found other services to be a bit more difficult to use.
And yes, I also agree with iFire's findings that another big difference is the number of models that are available. But the main differentiator with other similar services would definitely be the focus on its ease of use.
I should say that FinetuneDB seems like a solid competitor though!
I can say that for F# this has been mostly true up until quite recently. We use F# at work and were mostly unable to use agents like Claude Code up until the release of Opus 4.5, which seems to know F# quite well.
Thanks for the question and feedback. We are hoping to get some early testers on board at first. But yes, those are some good points, and we will be sure to add more details and perhaps some sort of a demo soon!
I'm excited to share Dashbullet, a new app that lets you build internal tools and dashboards simply by describing what you need in natural language. With Dashbullet, you can also interact with your tools and data conversationally, asking questions, updating information, or modifying dashboards just by chatting.
If you'd like early access, join the waitlist to be among the first to try it out. For regular updates on my progress, follow me on X (https://x.com/imjacquesmarais).