Jean-Baptiste Dézard retweetledi
Jean-Baptiste Dézard
6.3K posts
Jean-Baptiste Dézard
@jbdezard
Founder @DealExMachina. "Il n'est pas nécessaire de multiplier les choses essentielles." Science & Tech - Essential impacts on people & society.
France Katılım Mayıs 2009
3.2K Takip Edilen2.4K Takipçiler
Jean-Baptiste Dézard retweetledi
Ce soir en live sur @espacefm (21h30–23h)
Musique, émotions… et impact réel 💜
Fiers d’accompagner Lisa Collé dans ce projet 🐝
On compte sur vous !
@lisacolle_music instagram.com/p/DWOrz2MjAlW/…
Français
Agentic swarms need decentralized coordination. Some of the most compelling use cases of blockchains will be in this area.
Wei Dai@_weidai
Andrej Karpathy on autoresearch with an untrusted pool of workers: "My designs that incorporate an untrusted pool of workers (into autoresearch) actually look a little bit like a blockchain. Instead of blocks, you have commits, and these commits can build on each other and contain changes to the code as you're improving it. The proof of work is basically doing tons of experimentation to find the commits that work." The idea that distributed & permissionless autoresearch ~= proof-of-useful-work remains a high-level intuition for now, but it is extremely intriguing to say the least. Someone needs to take this further. See QT for more on what's missing.
English
@RaphaelDabadie @mcuban Workflows are preconstrained topological structures that manage an underlying state graph (lattice).
English
Exactly @mcuban.
The core issue is that it is incredibly hard to redesign workflows no one really sees.
Most of them live in people’s heads and have never been documented.
That’s why we’re building agents that first reconstruct the company’s operational graph, the workflows, dependencies, and ways work actually gets done across the organization.
Once that layer becomes visible, other agents can identify AI opportunities at scale, which is something no human team could ever do with the same breadth, depth, and speed.
Happy to talk about it
English
If my “repaving” you mean “reinventing “, yes. One of the challenges is that most corporate knowledge is still in someone’s head.
Knowledge is far different than information. LLMs and agents can capture all the information it can touch, internally and externally to the company
But there are things that you, me and everyone, security guards, salespeople, whoever, do to make the things we do fit the way we want them to.
None of that shit is documented anywhere. You can give LLMs or Agents to employees to improve how they do their jobs. But that isn’t all that much impact for bigger companies.
If you want to truly gain from AI, You can’t do it the way it was done, and just add AI. That’s why so many companies have struggled to get a return. Which I think is your point ?
You have to reinvent your business to maximize the benefit from the tech.
That is hard. There aren’t many CEOs capable of reinventing their company around tech they don’t understand.
Nor are there many shareholders in public companies willing to allow them to take that chance.
Let me rephrase that. There are ZERO public companies who will at this point, allow the ceo to wing it and reinvent the company around Ai. None.
It’s one reason @costplusdrugs has been able to reinvent our manufacturing. We optimize to the technology and train and retrain our employees to fit it. Our public competitors can’t begin to consider what we do.
You live this. Curious what you think about the above ?
Adam.GPT@TheRealAdamG
It feels like we are top of the 3rd inning. The models aren’t the problem, they’re smart enough now. Now it’s about applying them at scale. AI-enabling a process or workflow (like we’ve been doing) is one thing. But reimagining and repaving that process or workflow as AI-native is where transformational change will begin to occur — at scale. It goes slow until it goes really fast. I think that’ll be the story of 2026.
English
« A living operational graph of how the company actually works. »
Indeed. And agents should operate it, with HITL.
Bien vu!
Raphaël Dabadie (YC P26)@RaphaelDabadie
English
Jean-Baptiste Dézard retweetledi
En janvier 2026, l'artiste le plus écouté sur YouTube n'était ni Bad Bunny ni Taylor Swift.
C'était Slxughter, un producteur de phonk : 981 millions d'utilisateurs en un mois !
Tu connais ce son, tu ne savais peut-être pas que ça s'appelait "phonk".
blog.beewyze.com/phonk-bande-so…
Français
@ashpreetbedi Progress : Causal DAG x Bilattice x ISS.
Proven convergence.
English
@ashpreetbedi Seems to me that level 4 is a contentious territory : what’s the minimal scaleable non deterministic way to coordinate agents (governed and reaching finality) without pre wiring them. Not with Discord 🤗. Leaning on state graphs / events.
English
Jean-Baptiste Dézard retweetledi
Jean-Baptiste Dézard retweetledi
Can we replace orchestration-as-topology with governance-as-policy for agent coordination -- with formal convergence guarantees, bitemporal auditability, and a perpetual compliance lifecycle?
English
Darwinian — Marketing case — AI Native eco. Learn how to survive.
Anyone in the legacy tech should read and understand what’s coming. At which speed.
I’ve been a Cursor EA, and now I mix because the frontier between business and code has vanished.
Tanay Jaipuria@tanayj
Good piece on the "war time" at Cursor. Some interesting quotes: - The company’s new mandate was labeled “P0 #1”—priority zero: “Build the best coding model.” - Cursor estimated last year that a $200-per-month Claude Code subscription could use up to $2,000 in compute, suggesting significant subsidization by Anthropic. Today, that subsidization appears to be even more aggressive, with that $200 plan able to consume about $5,000 in compute, according to a different person who has seen analyses on the company’s compute spend patterns. forbes.com/sites/annatong…
English
Jean-Baptiste Dézard retweetledi
In large regulated orgs, we need governed state graphs for always-on agent swarms coordination, don't we ?
- Wu Wei style
- Least Action
- Margulis and lie groups, high dimensionality.
Like or follow if interested.
English
Jean-Baptiste Dézard retweetledi
A statement from Anthropic CEO, Dario Amodei, on our discussions with the Department of War.
anthropic.com/news/statement…
English
Jean-Baptiste Dézard retweetledi
Jean-Baptiste Dézard retweetledi
Why I'm excited about RLMs: they're a simple, generally applicable test-time strategy with tons of low-hanging fruit for optimization.
Great for dealing with long contexts today. Tomorrow? Could be so much more... dbreunig.com/2026/02/09/the…
English
« Contact sales », c’est l’abandon de panier assuré. Le tue l’amour.
On a Stripe aujourd’hui, X402 demain.
Vous avez combien de « CTA » qui tuent votre revenu passif?
Cc @Maldoju
L3 Tweet Engineer@MegaBasedChad
>be me >want to pay @MistralAI money for their cool new audio transcription API >API docs have a "contact sales" button instead of a stripe link >Just run the model locally instead because I don't feel like fussing with a sales rep Do Europeans hate making money?
Français
Jean-Baptiste Dézard retweetledi
The world loses millions of dollars in productivity every day that coding agents aren't RLMs yet.
English
The long view.
The willingness to act forced clarity about the capacity to act.
geopolitique.eu/en/2026/02/02/…
English