Elliot Vaucher (@ElliotVaucher) header
Elliot Vaucher (@ElliotVaucher) profile picture

❤️ Intelligence : artificial or not. Solving the AI onboarding and customisation problem for elite independent law firms and finance actors.

Lausanne, Switzerland
Joined November 2018
Photo by Elliot Vaucher (@ElliotVaucher)Photo by Elliot Vaucher (@ElliotVaucher)Photo by Elliot Vaucher (@ElliotVaucher)Photo by Elliot Vaucher (@ElliotVaucher)Photo by Elliot Vaucher (@ElliotVaucher)Photo by Elliot Vaucher (@ElliotVaucher)Photo by Elliot Vaucher (@ElliotVaucher)Photo by Elliot Vaucher (@ElliotVaucher)Photo by Elliot Vaucher (@ElliotVaucher)Photo by Elliot Vaucher (@ElliotVaucher)
Best proxy for good hires in AI is how long you can stay away from any intelligence optimisation task
2
432
Le pari de pascal version AI. C’est possible qu’on assiste à la plus grande révolution technique qu’on connaitra de notre vivant. Oui, ou non. Si oui, qu’on y croit et donc qu’on en profite chaque seconde, on gagne tout. Si oui, et qu’on n’y croit pas et donc on s’en distancie, on rate probablement les découvertes des prochaines décennies en temps réel sous nos yeux. Si non, et qu’on y croit quand même, on assiste probablement à des choses assez spectaculaires, sans être transcendantes, divertissantes tout au plus. Si non, et qu’on n’y croit pas, on continue notre vie comme avant et on ne perd ni ne gagne rien. Dans l’absolu, et toutes choses égales par ailleurs, bien plus de bénéfices à y croire.
44
at this point, why do you even bother explaining yourself about why timing is unfortunate (oh no, there's an IPO going on). posterity will need another version expunged from these trivialities, they're just embarassing alpo.ge/remark.pdf
10
Looking forward to the coming (stupid) debate on vibe solving and vibe discovering (mathematics, biology)
8
How quickly things change. One day X has the reputation of a dead social media, the next it’s the place where you read daily about historical scientific and mathematical breakthroughs while your standard newspaper still posts about “industry not seeing ROI on AI”.
4
Imagine being Thélyson Orélien and trying to explain to the Goncourt jury this is how you worked.
an MIT researcher built a realtime LLM vector steering tool and it’s the coolest thing I’ve seen someone build in a very long time. this is so brilliant. the primary use case I see this realistically used for is long form writing tasks. some might find it controversial, considering it plays into llm-generated writing. however in my opinion it kind of solves for some amount of the tension around the issue by making the process multiplayer in a new way. the human and the AI become collaborators, sort of. idk this sh*t is so cool though, man. (@ far__behind on instagram)
1
30
fundamental dilemma for OpenAI is control vs ease of use. even in the eventuality of auto optimisation on client context, the problem stays the same. Ease of use (auto configuration) diminishes control (in terms of pinpoint accuracy of controlling parameters of a tool). the question is not exactly is the tool powerful, but how much can we control it. Cause there is power in control.
12
Today’s dotcoms are all the companies trying to disintermediate experts.
2
23
Finally a rework of the original excel elevator commercial
Some of you have been asking very politely. Some of you have just been asking. GPT-6.2 Astra and Sol are now available. Super efficient and built to perform where it counts—the real world. openai.com/index/introducing…
1
29
The only program that recursively auto optimises with the perfect equilibrium of conservation and exploration inside the context of the real world is life. End of story. Trying to recreate synthetic life through science is the never ending quest of humanity.
2
29
Helicopter money in 2026
you can hide bitcoin inside a song. press sing on a laptop and it plays a jingle. your phone hears it, extracts the ecash from the song, and a that bitcoin is redeemed to your lightning or ecash wallet. it's called hush money, and it's live on magical.cash right now.
1
1
83
end of billable hour for law firms in the era of AI is a false problem big law leverage currently sits on 2000$/hour partner using 500$/hour associates to do the job the spread is the leverage with AI it's just even bigger (50$/hour of tokens instead of 500$ for associate) if clients are not willing to pay 2000$/hour for senior partner who offloads to Claude, this initial rate will lower but the spread will keep existing am I missing something ?
39
Breaking news : turns out Jürgen did not invent NVIDIA chips before NVIDIA.
Congrats to @NVIDIA, the first public $4T company! Today, compute is 100000x cheaper, and $NVDA 4000x more valuable than in the 1990s when we worked on unleashing the true potential of neural networks. Thanks to Jensen Huang (see image) for generously funding our research 🚀
31
Elliot Vaucher retweeted
Reality of our days: 1. I am sitting on a set of vulns (4+) that allows me to remotely take over a consumer home automation device I have (3+ diff ways). 2. The vendor doesn't only serve consumers, but also critical infra (announced partnerships include pharma manufacturing plants, airport logistics, hospitals, etc) across various locations in Europe and the US (not the same product line, ofc, so maybe moot). 3. Claude and Astra refuse to help me explore even my own device, so now I'm using Chinese hosted models for investigating. 4. The vendor (with a revenue of $0.75B) does not maintain any security disclosure practice whatsoever, nor does respond to emails. So, to recap, I need to outsource vulnerability hunting of (potentially, I don't know) critical western infrastructure to Chinese models, because the western models refuse; and the vendor doesn't give a shit. Fun timeline. PS: The vendor is Swiss.
126
290
52
3,822
177,009
There is a strange but essential economic effect at play in this AI revolution. The harder the evaluation of the quality of the outcome of a model becomes for a human being the higher the importance of emotional loyalty to a brand. This defines a gap between objective and perceived quality, where a lot of actors will be able to place products.
1
19
We should run Opus on Hegel, we might finally be able to go further than the introduction.
20
We just moved engineering one level up the computing stack. 1940-2000 : hardware era 2000-2023 : software era 2023 onward : flexware ? (lame) aware ? (funny but…) When building on top of llm based platforms like ChatGPT (codex powered desktop app + complete OAI api and services), the complexity and precision of engineering doesn’t magically disappear, it becomes a matter of how well do you formulate an objective, how efficiently do you allocate compute to solve it, how do you create algorithms that auto solve, auto improve, and so on.
Brute force computation is not enough for intelligence. Clever new methods are needed. The science of AI is far from finished. IMHO
2
113
The sweet lesson
Brute force computation is not enough for intelligence. Clever new methods are needed. The science of AI is far from finished. IMHO
23
Paradoxically valuable proprietary data is a moat that is going to require a lot of ressources to secure in the coming reality. Is a defense that needs to be defended still one ? Logically, no. End of story.
19
At this point I’m really taking it as a compliment that startup competitions jurys don’t understand what we’re building with ogram. We’re radically innovating. Technology, business model, pricing. The sheer amount of demand, and our clients speak for themselves. The need exists.
16