Carlo Daffara retweeted
I'm obviously biased since this is my song/video, but I think what you can do with AI now rivals any large studio or production. This one was made with Grok, Suno and Runway.
The videogen cost ~$100 in tokens; I'm certain that if I spent a full day editing (instead of ~2.5hrs) and used a budget of $250-$500 I would produce a video so realistic that no one would be able to tell was AI. A lot of it is in the edit - make no mistake, these songs/videos do take real human work, but if you have a vision for the end result you can bring it to life without constraints. It's awesome. x.com/marksoares/sta…
The videogen cost ~$100 in tokens; I'm certain that if I spent a full day editing (instead of ~2.5hrs) and used a budget of $250-$500 I would produce a video so realistic that no one would be able to tell was AI. A lot of it is in the edit - make no mistake, these songs/videos do take real human work, but if you have a vision for the end result you can bring it to life without constraints. It's awesome. x.com/marksoares/sta…
32
15
178
7,952
Carlo Daffara retweeted
Companies that offer integrated solutions for local inference will boom in the next few years. I'm not talking about another GPU in a case.
Hardware + OS + an inference stack that makes open models easy to run. A plug-and-play system.
Ease of use is still underrated.
Hardware + OS + an inference stack that makes open models easy to run. A plug-and-play system.
Ease of use is still underrated.
123
97
1,699
169,370
Expect this more and more.
BREAKING: Revolut moved ~2,000 engineers onto self-hosted GLM for coding and pushed Claude Code and Codex to the exception list.
Per David Abitbol, who leads AI ops there: 56 Nvidia B200s, a blended 30 cents per million tokens, roughly a tenth of what they paid Anthropic and d time improved 70%.
Here's the part that matters: the gains showed up when tokens stopped being rationed. Engineers quit saving AI for the hard tickets and pointed it at everything.
Frontier models become the exception, open weights become the default. That's a pricing problem for every lab heading to IPO.
h/t @OliverMolander
Would you move your team off frontier APIs if the bill dropped 90%?
0
0
1
18
beautiful.
This is one of the most mind-blowing stop-motion films we've seen this year. It comes from India.
Using traditional ceramics, Studio Eeksaurus invented a new style of animation. We're exploring:
─➤animationobsessive.substack.com/p/animating-tr…
0
0
1
55
Carlo Daffara retweeted
LISTEN VERY CLOSELY TO WHAT $PLTR IS SAYING:
"L3 Harris noted that by fine-tuning open-source models on its own data, a sovereign model outperformed the frontier model it had been using in less than 48 hours and cut costs by 95% in preliminary results. $NVDA reported that by evaluating and post-training its open models, it was able to get its smallest most efficient model to outperform its largest model by a large gap, and can now retrain that model on just a single GPU in a couple of hours, enabling them to retrain any time they have data to improve the decision intelligence."
"Tampa General is deploying sovereign AI to keep sensitive data and model processing local while enforcing its own risk and privacy standards for LLM traffic and supporting the granular controls Tampa General wants to define for itself. Using fine-tuned and post-trained Nematron to achieve this outcome, together we also improve model performance by 7% and reduce latency by 22x."
"L3 Harris noted that by fine-tuning open-source models on its own data, a sovereign model outperformed the frontier model it had been using in less than 48 hours and cut costs by 95% in preliminary results. $NVDA reported that by evaluating and post-training its open models, it was able to get its smallest most efficient model to outperform its largest model by a large gap, and can now retrain that model on just a single GPU in a couple of hours, enabling them to retrain any time they have data to improve the decision intelligence."
"Tampa General is deploying sovereign AI to keep sensitive data and model processing local while enforcing its own risk and privacy standards for LLM traffic and supporting the granular controls Tampa General wants to define for itself. Using fine-tuned and post-trained Nematron to achieve this outcome, together we also improve model performance by 7% and reduce latency by 22x."
45
114
1,156
312,515
It makes pretty sense if they think they can reach full R&D automation and AGI first.
If I'm not mistaken, Anthropic's 85% gross margin and >40% compute margin from this Bloomberg story implies that 70-75% of their compute is now going to R&D.
bloomberg.com/news/articles/…
0
0
4
50
Carlo Daffara retweeted
Meet Mistral Large 4, aka Le Chonk.
• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.
Open weights release end of October.
• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.
Open weights release end of October.
2,035
4,561
40,603
5,763,527
Carlo Daffara retweeted
Claude 5.5 Opus responding to “AI is a normal technology” is still the best AI-created video I have ever seen
212
809
5,334
1,109,313
Carlo Daffara retweeted
i gave my ai agent (fig running on opus 5.5) access to the @runwayml MCP and using seedance 2.5, nanobanana pro, lyria & hyperframes it made this
you might've seen the chunks i posted over the weekend but here's the full 27m doc
i only gave notes like a producer would
you might've seen the chunks i posted over the weekend but here's the full 27m doc
i only gave notes like a producer would
22
20
200
13,328
Carlo Daffara retweeted
Made with Claude Opus 5.5.
The fearmongering about AI always makes us forget that NOTHING WENT FOOM, as was always predicted.
Send this to your doomer friend who has a very high P(Doom).
Accelerate.
The fearmongering about AI always makes us forget that NOTHING WENT FOOM, as was always predicted.
Send this to your doomer friend who has a very high P(Doom).
Accelerate.
479
873
5,769
909,427
Carlo Daffara retweeted
Today I'll present Boundaryless #O2A (Open Organization Alliance) protocol at the Protocol Symposium 2026 from Protocol Institute.
Livestream youtube.com/watch?v=ddUfKj….
The O2A is an open language to describe company world models as nodes, offerings, contracts and their compositions: it's an #opensource language that you can adopt to describe and program your organization, and it supports the convergence in operating models we're seeing since a while now - that I discussed recently in my article: "TTB 4: Everybody Is Saying the Same Thing in Organizational Development" lnkd.in/dVK3JqTt
We're about to organize a presentation webinars and now collecting interest for organizations to co-author this standard so reach out if you're curious on how this can help you integrate #AI agents, and cooperate more easily internally and externally!
Livestream youtube.com/watch?v=ddUfKj….
The O2A is an open language to describe company world models as nodes, offerings, contracts and their compositions: it's an #opensource language that you can adopt to describe and program your organization, and it supports the convergence in operating models we're seeing since a while now - that I discussed recently in my article: "TTB 4: Everybody Is Saying the Same Thing in Organizational Development" lnkd.in/dVK3JqTt
We're about to organize a presentation webinars and now collecting interest for organizations to co-author this standard so reach out if you're curious on how this can help you integrate #AI agents, and cooperate more easily internally and externally!
0
3
3
289
Apart from a few nitpicks (the chess board with... too many pieces) this is excellent. Love the dark humor.
i'm still not sure people are fully getting how good opus 5.5
i gave my claude agent fig access to the runway MCP and told it to make a 'high-end netflix style documentary about superintelligence for normies' and...
it came back with this
0
0
7
69
AI knows how to assembly an IKEA cabinet. We are doomed.
Can AI tell if you've built your IKEA furniture wrong? Our new benchmark, the Furniture Assembly Benchmark (FAB), gives models the manual and a photo of a half-completed piece of furniture and asks them to spot the mistake.
The top score has gone from 28% to 80% in just 10 months.
0
0
0
47