Merlin
1.1K posts
Merlin
@Merliiin__
Building projects and study AI I look for opportunities, I find opportunities
Heaven Katılım Şubat 2023
300 Takip Edilen2.7K Takipçiler
GPT-5.6 Luna deserves far more attention
much smarter than 5.5, not a yes-man, and will actually correct you, and a major reason is that 5.6 Sol autonomously post-trained it
so it is not just a smaller model, but its training was improved by a more powerful model, not only by human researchers
English
@VadimStrizheus The best approach is to work using the Fable + GPT-5.6 combination
English
@ivanfioravanti Grok has performed excellently, so it should handle the tasks well
English
Ok, let's push Cursor + Grok 4.5 on something really complex. Metal Kernel for ds4 optimization. 😎
English
Your AI subscription resets every week. Your brain doesn't come with that feature
It's the weekend - the perfect time to take a break from social media and technology
Let your brain simply live and enjoy the world around you
AI limits get refreshed all the time, but your mental capacity doesn't work that way
So give yourself a rest and go touch some grass
English
you keep telling me hugging face would never. cool. anthropic would never take a model dark either, then the government made them do it for 19 days and nobody asked hugging face's opinion.
"open" is a license, not a guarantee. the only guarantee is a file on a drive you control.
Sudo su@sudoingX
the entire open ai movement lives on one website. every model you think you have access to, qwen, gemma, deepseek, llama, all of it sits on servers owned by one venture funded company. one acquisition, one export control order, one licensing policy shift, and the door closes overnight. it already happened to a frontier model three weeks ago, nineteen days dark by government order, and nobody saw it coming. i'm not predicting hugging face disappears. i'm telling you that you are one board meeting away from finding out how much of "open" ai you actually own. weights on your drive are the only ai you own. pull them tonight.
English
@Merliiin__ Hmmm that is true, but personally I dislike a lot of the releases. Ghost in the Shell is exactly my jam, I love sci-fi and cyberpunk. We don't get enough of that.
English
There's a new "Ghost in the Shell" anime on Prime Video! It has a very different vibe and visual style from the earlier shows and movies, and feels much closer to the original manga.
Only one episode is out so far, but I loved it.
Have you watched it yet?
English
@DavidOndrej1 You seem like an OpenAI hater for now - are there specific reasons why you dislike GPT so much?
English
@DavidOndrej1 Generally speaking, whenever a new AI is released, everyone writes glowing reviews - regardless of whether it’s Fable, Grok, or GPT
English
does OpenAI have half of the AI influencers on payroll?
because I'm not impressed with GPT 5.6 at all...
English
Anthropic's "coincidences" this week aren't coincidences
First, Anthropic quietly extends Fable 5's discounted credit window - right as GPT-5.6 rumors start heating up
Then, the moment GPT-5.6 gets its wide public release, Anthropic resets 5-hour and weekly limits for every Claude Code user
No explanation given. Two moves.
Both timed to the day a competitor makes noise. Maybe it's just infrastructure. Maybe it's a defensive playbook. Either way, the pattern is hard to unsee
English
Realistic review of GPT-5.6:
> Sol is a bit worse than Fable
> Sol is seriously slow and drains your usage limit 5x faster
> Terra is basically the same as 5.5
Summary: GPT-5.5 is still here as Terra, and the update just introduced a better, slow, expensive subscription option called Sol. (Api price is still same)
English
@VadimStrizheus @thsottiaux Does that kind of check consume a lot of tokens?
English
I just asked GPT 5.6 Sol Ultra to review my app, and scan for vulnerabilities and its doing it.
CODEX IS SO BACK
@thsottiaux
English
@DavidOndrej1 We’re waiting for details on why you disliked it so much
English
this is worse than i expected
David Ondrej@DavidOndrej1
first impressions of GPT 5.6 are not good...
English
GPT-5.6 one-shotted a complete 3D Manhattan build after running autonomously for nearly a week.
The output reportedly includes accurate building placement, street layout, and city-level detail.
Matt Shumer@mattshumer_
GPT-5.6-Sol one-shotted this voxel-based Manhattan. Just look at the precision... it's insane. It ran for almost a week, completely autonomously, to get the job done.
English
@DavidOndrej1 Do the limits really get used up that quickly now? Does that mean GPT is no longer more cost-effective than Fable?
English
the $200/mo Codex plan used to be a steal
I never hit the limits, even running multiple /goal loops with GPT 5.5 xhigh
I could literally have 10+ agents running in parallel and would NOT hit the limit. ever.
but now, with GPT 5.6 Sol, i hit the limits instantly
even while running just 1-2 agents
i'm not happy with this at all...
English