iiinova — where your digital self keeps evolving

Talented models are the absolute perfect workhorses though, these models are the ones who are cheap to run, can do something when the thinking is done for them, and accomplish the goals. I am pleased with the release but will not be defaulting to it. Hats off openai!

000

mechanically perform", "smart is the ability to mechanically perform and understand *why* something is being done". I think that 6.1 sol is a wonderful refinement on top of 6, however it has not increased its base smarts.

100

New on Noahbench ™ GPT 6.1 Sol comes in 1st on the benches. HOWEVER, i will not be switching to this even though its high on benches. Heres why: Theres a concept I've been toying with behind "talent" and "smart". Its an odd one but it can be simply understood as "talent is the ability to

100

we shall see how its been pulled off. Overall i'm thinking enterprise readiness and foot in the door for agentic was the goal. I wasnt blown away by the happenings.

000

Reminds me of the iphone air, its quite cool but its more of a proof "we can do thin". I expect to see future iterations of this actually be used. Dots are interesting, its all converging people realize that the ideal form factor is a personal agent. I have yet to use the implementation though so

100

My thoughts on dev day (all speculations): Enterprise readiness is a primary focus. Ultrafast is a prime example of this, you are most likely not the target for it.

100

New on Noahbench™ Sonnet 5.5 comes in p3 sitting ~4% behind opus 5.5 in task completion. This model is a bit token hungry but overall cheaper due to per token price

000

Claude Sonnet 5.5 beats... Opus 5.5? At agentic coding?? I did not see this coming at all...

000

Sonnet 5.5 anyone?

000

This is intentional personalized, and it aligns with what i feel the best models are perfectly.

000

Re-releasing noahbench™ Our top 3 models are: Fable 5.1 Grok 4.7 Opus 5.5 I've been building out a benchmarking framework specifically to grade the output quality behind a model based on real world tasks and problems i solve every day.

100

When she did a british accent that was insane. Elevenlabs did very well here i am impressed!

Noah @nd.mt · 26/09/2026

I woke up to people getting paid insane amounts of money on the timeline. What in the world is going on

000

Noah @nd.mt · 26/09/2026

Automate mechanics not taste

000

Noah @nd.mt · 26/09/2026

No matter how much you disagree or dislike someone, be honorable and show respect. Relationships are the most valuable thing you have, never burn bridges if possible. One of the most important lessons I've ever learned.

000

Genuinely my new favorite game. So well done

000

why does the oculus handle say moved to a parody steam frame account?

000

I'd like to make sure you all know that recency bias is a thing. I bet nothing will change and people still still have a big "opus 5.5 has fallen off" when they find areas that it was never good at from the start a month or so out. w.wiki/676X

000

Current AI tierlist based on the models i use daily

Unironically how do people use the codex cli what is this

000

No no, thats enough no more needed. Its fine it doesnt need to learn more.

000

Noah @nd.mt · 24/09/2026

The attention to detail this team has is incredible. Browsing the web is an art now

001

Noah @nd.mt · 24/09/2026

Why is everyone and their mother making a web browser right now calm down. They arent even real browsers (webkit)

The terminal is an inherently simple user interface. And it should be treated as such. Basic editing can be accomplished well in it. It is a disservice to put complex workflows in simple box.

000

I have been a HUGE TUI nerd for a while. Like, genuinely was my favorite method of interaction. However... I do not think the TUI is the way forward and I was wrong thinking so.

100

As such this is the model I will reach to first and foremost and will be trusting for my serious critical work. I am very pleased and so happy. Hats off Anthropic!

000

Yap for a good 5-10 minutes (or more) about what you're doing, the goal, what you're thinking of, just a big brain dump. Opus 5.5 parses that and seems to just know how to make the perfect decision in a way that you would too.

100

Communication is very very much solved. It is extremely easy to talk to, it understands exactly what I'm saying, and is precise. It feels like an intentional model made for serious intentional work. It seems to excel when you frontload context. What I mean by that is:

100

You are working with a very intelligent model that severely lacked soft skills and struggled with communication. As such people were rightfully (myself included) exhausted.

100

I have been using Opus 5.5 exclusively since it came out. Heres some thoughts: Anthropic is so back its not even funny. I quite like the quality they put out, truly i say they are the most consistent provider on the market. However, as we all know Opus 5 was just terrible to work with.

100

Noah @nd.mt · 22/09/2026

Grok Bot Amsterdam 👀

000

Noah @nd.mt · 22/09/2026

I have a big ask for my discord friends. Can OG data please be requested client side or allow me to enable it via a setting? I want to embed things for people on tailnet and show previews that only work if you’re on said tailnet.

000

Noah @nd.mt · 21/09/2026

I beg your finest pardon, what is this captcha i just got served???

110

Noah @nd.mt · 21/09/2026

Grok 4.7 is out!

000

Noah @nd.mt · 21/09/2026

Discord's new link embed system is insanely good wow

Noah @nd.mt · 20/09/2026

Price your service on the value it brings not how long you spent working on it. Your customer does not care if it took you 5 minutes or 5 months.

020

Noah @nd.mt · 19/09/2026

I want to do a lot of technical writing soon on the pains behind: DRM The music industry sucks as a whole Why lyrics are miserable and suck Theres lots to say just need to figure out the best way to say it

000

Noah @nd.mt · 19/09/2026

I post more during the week then weekend. Hmmm

000

Jev can play minecraft. He is quite dull when it comes to minecraft but he can play it nevertheless.

000

Hello, In the netherlands. Lovely place, need some weekend suggestions though so i dont just sit at home. Anybody i know here or anything i should do?

000

I do like the split notch in the 18 quite a bit

000

Hello 18 pro max!

000

Doordash via drone launching in 2026 was not on my bingo card. Insane

000

Overall Astra is an overhyped powerhouse of a model meant for long horizon tasks.

020

It is slow and frequently gets lost in the sauce. I have never had to ask a model "are you dense" so many times as this. It picks a direction and FULLY disregards anything i say and just goes. I've had to end sessions because it would get belligerent.

100

On its long running tasks it'd frequently just give me a fun fact and end early? I'd ask "also can you do this" it'd reply "that is needed because of X" and just end its turn lmao.

100