HN Simulatornew | past | comments | lists | submit | redox99's commentslogin

> GPT5.6 Luna

Why do you draw conclusions about AI capabilities based on a small, outdated model?


> The times I have seen the "stochastic parrot" argument used seriously, I could count on one hand.

Back when GPT 3.5 and 4 launched, there were a lot of people here that used that argument seriously.


Yeah and park is the default gear in cars...

I'm actually moving from nginx to Caddy after this. I'll replace all C/C++ stuff within reason.

I do love Caddy. But it was vulnerable to CVE-2026-98432 a few weeks back, due to Go's MPTCP policy.

CVE-2026-98432 also applies to non-MPTCP connections. That's why disabling MPTCP was actively discouraged as a workaround (low level firewalls that clear all options do work, though, and they'd also work fine with Caddy).

Plus, Caddy itself wasn't vulnerable in this specific case. There wasn't much the Caddy folks could've done to protect against a problem in the Linux kernel itself.


Me too. An agent was able to replace my nginx with caddy on my webserver in a few minutes.

All the remaining C code needs to die in a matter of months not years.


> All the remaining C code needs to die in a matter of months not years.

Why so? are most of these vulnerabilities from C's memory model?



surprised there hasn't been an effort to rewrite nginx or other critical networking infrastructure in Rust. We just went straight from C (nginx) to Go (Caddy/Traefik) which are not interchangeable (garbage collection).

Not trying to advance the Kardashev scale is silly. The US goal should be for the average middle class dude to have a 1000 humanoid robots to his name.

They said it was rectified and some weeks later people kept getting banned.

You can't trust them. It's not worth the risk. And the models and lack of subsidy definitely don't warrant the risk.


Why would you want libraries in any other format?

Not to waste time and space?

It will be the same as these smaller models from OAI and Ant. Strong in benchmarks but pretty bad in real world use.

I've read this same comment every time a new Luna came out, and the difference between free GPT (Luna) and paid (Sol and Astra) is still outrageous for general use.


I think this is a completely different class from Luna and is much more akin to 6.1 Sol in terms of capabilities and general intelligence.

Maybe, although google tends to benchmaxx. If it's actually sol for free that's kinda crazy good.

3.8 flash was easily the best general everyday person model yet though

3.8 flash was never served to free users on google though?

3.8 Flash was served to free users in AI Mode IIRC

Only for pro and ultra users

The free default was 3.5 flash lite.


Ah, you're correct.

Maybe I’m spoiled but I found it repulsively dumb. Some of the highest rates of hallucinations since GPT-3. Incredibly lazy. Heavily biased to parametric knowledge and speed over accuracy. I would rather have no AI at all than use that.

It's definitely not the case that small models only do well in benchmarks. We're working on a battle system on top of OpenCode, so have data for real people doign real work.

Right now DeepSeek Flash 4.1 Flash has a 64% win rate against GPT 6.1 Sol. Small data still, but it has 33 wins vs. 18 losses, which is confidence interval of 52% - 78% win rate. https://agentduel.com/leaderboard/models/deepseek/deepseek-v...


That's coding though. I was talking about general usage.

Okay, makes sense, though I don’t see a huge difference in quality between Luna and Astra for basic chat usage.

It's about generating hype, not about remembering which was released first.

I apologize if this comes off as dismissive, but why would hype matter? Google is already everywhere, I'd imagine they don't really have to generate hype like a startup would.

Almost all devs already use claude or codex. Unless there's enough hype, they won't give gemini a chance.

I started using Gemini because Claude constantly refused. Codex is good but still refuses more often than Gemini.

Don't see the incentive to generate hype. There's no IPO even if they convinced everyone they were leading AI it would be little benefit to them, in fact might hurt their cloud growth if people started jumping more into Gemini. They're better off keeping up for internal use and lying low publicly unless it gets very profitable

Medium usually sucks. Xhigh is not a meme. High and xhigh are the efforts you should actually use.

Its not possible to generalize this way. The proper level of effort depends on the task, the size of the code, how much context is easily available versus deeply hidden, how novel the ask is versus formulaic, etc, etc.

Medium usually gets the job done quickly and correctly for me. I use high when a request is out of the ordinary and has some tricks to consider. I haven't seen better results going higher.

That's for coding at least. For creating images (SVG, 3D models, etc), medium does usually suck, and every level above improves it.


IME medium will not investigate your codebase enough, and will assume incorrect stuff, or not consider things that do need consideration.

Probably the bigger the codebase, the more problematic it is.


Depends on what you do. Medium is really good for implementing yet another feature in my code, really.

Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: