Maximilien Pressensé

AI-pilled motherfucker.

[Maximilien Pressensé @MaxPressense]

API token wise I mean.

[Maximilien Pressensé @MaxPressense]

I feel the current AI meta is off atm.

Opus 5.5 is impressive and deserve all the praise. It's SOTA on raw intelligence.

But GPT 6.1 Sol is the most useful one.

[Maximilien Pressensé @MaxPressense]

@lydiahallie I've been running it for both codex and claude for a couple weeks (my own implementation). Great feature indeed.

[Maximilien Pressensé @MaxPressense]

@samruby But those projects did get more attention after rails world, don't you think ? I figure you'd have the data ...

Anyway, truth < storytelling.

[Maximilien Pressensé @MaxPressense]

@daedalium Makes me want to watch this classic again.

[Maximilien Pressensé @MaxPressense]

@doodlestein I thought the same about getting Grok to the party, a month ago. Defi not ready at the time.

Tbh I don't believe the frontier club has more than 2 participants still.

But soon they could be 5. I'm rooting for that too !

[Maximilien Pressensé @MaxPressense]

I feel the current AI meta is off atm.

Opus 5.5 is impressive and deserve all the praise. It's SOTA on raw intelligence.

But GPT 6.1 Sol is the most useful one.

[Maximilien Pressensé @MaxPressense]

Unless you have infinite money to buy infinite amount of subs that in return pay for themselves, then go on, just use Opus 5.5 for everything.
It is the best model. The real frontier.

But if you are from the 99% of the 0.2% using models for a living, just put gpt 6.1 Sol high in the mix.

It is the only fundational one for fully automatic agentic workflows.

It's cost/intelligence ratio is unmatched. People and Businesses need to start acknownledging that.

No lessons to be drawn from Theo's usage imho, quite the opposite.

[Maximilien Pressensé @MaxPressense]

Unless you have infinite money to buy infinite amount of subs that in return pay for themselves, then go on, just use Opus 5.5 for everything.
It is the best model. The real frontier.

But if you are from the 99% of the 0.2% using models for a living, just put gpt 6.1 Sol high in the mix.

It is the only fundational one for fully automatic agentic workflows.

It's cost/intelligence ratio is unmatched. People and Businesses need to start acknownledging that.

No lessons to be drawn from Theo's usage imho, quite the opposite.

[Maximilien Pressensé @MaxPressense]

@Phymon @ylecun @elonmusk @zerohedge Yes. I guess you didn't get the joke then.

[Maximilien Pressensé @MaxPressense]

@ylecun @elonmusk @zerohedge Time for Meta AI to get a new leader ... oupsy :p.

[Maximilien Pressensé @MaxPressense]

@thsottiaux I was confused too.

I guess not all you customers feel the pain of the weekly limit.

[Maximilien Pressensé @MaxPressense]

Everybody restarted as a junior against this paradigm shift.
We're all one-year olds.
Many out-of-tech workers aren't even born yet :p.

How does the junior person learn to become the senior person? It's about teaching them the proper mindset, as always, and giving them the proper context to ingest.

If you mean 'better' by 'senior' (it always have been a weird assumption on the working place, as experience doesn't usually translate in better problem-solving skills, nor higher agency; senior just used to know more stuff, which is an advantage that AI is about to erased), then mindset is the ultimate differentiator: curiosity, high agency, willingness to help, passion, self-awareness.

I believe this is what you've been teaching all along through your Family's work.

How do you learn the proper mindset ? peers, role-models. Therefore, schools like 42 aren't going to be out-of-business anytime soon, they're actually more relevant than ever.

[Maximilien Pressensé @MaxPressense]

Everyone is going to be a creator.

There will be art for everybody, every nich, every fetichism.

This video has 80's vibe, when no movie was ever made without selling sex. It was just the meta of the time I guess.
Then the meta evolved. Nobody could do this meta mainstream thereafter without being cancelled.
Now, everyone is free, and we'll see what people really want.

The video also demonstrates that making art needs dedication, taste. This clip is rubbish, it's just an interesting demo on this tech SOTA.

Special talent will still exist.

Isolating signal over noise will be more on more important. I believe curators, as influencers will keep being a big part of our lives.

[Charles Curran @charliebcurran]

If you think AI film can’t be art then explain this.

[Maximilien Pressensé @MaxPressense]

actually I will not, I don't want the ads, no thank you.

Funny how Brave + Youtube on the mobile is actually a free premium youtube sub.

(and even better tbh, as the algorithm of refreshing your feed doesn't work as well on Brave and therefore do not propose you new content all the time, sucking your soul out, you mere dopamine addict)

[Maximilien Pressensé @MaxPressense]

Now I might use Youtube App again, instead of the website on Brave with native anti-shorts filter.

[TeamYouTube @TeamYouTube]

@petergyang You have the option to set your Shorts time limit to zero. You shouldn’t see Shorts on your YouTube homepage anymore if you set your time limit to zero. More info on how to set your limit here: https://t.co/Ac14SgQUdk

[Maximilien Pressensé @MaxPressense]

What's the business model of T3 code ? I see it's OSS and that's great, but I'm pretty sure you've got a team behind it that you need to pay ...

I guess you might already have shared your views, can you (or someone passing by) link me to it ? (not a video please, a text has to do)

I've been vibe-coding my own meta-harness for a year, juggling with 2to3 max subs from different AI provider, playing with Software Factory pattern, Orchestration pattern (CoS), role based agents.

And I discover just now that T3code was solving the same problem all along. Or a good part of it.

I don't regret a single bit spending almost all my time tinkering, and learning AI applied engineering through this, but I'm wondering If I should now invest on T3 code.

[Maximilien Pressensé @MaxPressense]

@florin_dev @poteto This is wrong man.

Unless you want to be a tech influencer though ^^

[Maximilien Pressensé @MaxPressense]

This is not a secret, but it is one of the biggest differenciator between top performers (any intellectual job remaining :p) and the masses.

Fight the cognitive lazyness !
You are responsible for your understanding. Do your own work !

So many people ask for meetings, video tutorials, etc because they just can't bear the load.

Brain is a muscle. Train it, or be replaced.

[shadcn @shadcn]

The most important coding skill: read.

Teach yourself how to read.

Read the agent output. Read the blog post. Read the paper. Read the code (if it applies to you).

Don't summarize.

[Maximilien Pressensé @MaxPressense]

As a freelancer, I felt this risk early on.

My 'normal job' is now only the consequences of my experiments with AI.

Side effect, I got much more done on my 'normal job' by not focusing on it primarily.

But there's variance to it.
Tinkering means breaking things.

On the lg though it benefits both me and my employer.
But alas, my future ex-manager doesn't see it this way :p.

[geoff @GeoffreyHuntley]

if your team is too busy doing their 'normal job' to experiment with AI, you're preparing them to be replaced

[Maximilien Pressensé @MaxPressense]

@theo Did it ?

[GM @thisisgm]

There should be a "donate the rest of my usage for the week to open source" button in Claude and Codex. Imagine your reset hits in 12h and you know you won't use it. Click the button, go to bed, and you just gave a poor maintainer halfway across the world a lifeline.

[Maximilien Pressensé @MaxPressense]

@dexhorthy What do mean about the hard learning almost two years ago ?

Are you sure it still applies ?