Sacha Martinelle

Building infinite worlds. Prev. @theresidency

[Sacha Martinelle @chasnel_m]

I’m launching https://t.co/TTBUEsLo9P with @JuliusCasio to make indoor cycling 100x more immersive.

We’re at @joinswanbase and looking for beta testers to rapidly improve the product.

Come join us in the 2ème arrondissement!

[Sacha Martinelle @chasnel_m]

@DFintelligence Le substrat glisse des données textuelles vers le monde physique, mais c’est toujours l’intelligence qui créée la majeure partie de la valorisation. Les superlevées en robotique humanoïde sont un pari sur la valeur des modèles, donc du software. Pas du hardware.

[Sacha Martinelle @chasnel_m]

Very accurate. I decided to go deep into sales and coding at the same time, a year ago, because I couldn’t imagine being a builder if I couldn’t convince other people to acquire what I build.

[Brian Cho @brianyoungilcho]

If you can’t sell, don’t start a company.

Yesterday, @thejessezhang came to the @a16z @speedrun office to share what he learned building @DecagonAI. The biggest point that resonated with me was simple: The founder is the best salesperson a company will ever have.

A for-profit company does not become a real company when you build a product. It becomes a real company when someone pays you for it. Yet so many founders (especially developers-turned-founders in SF) do the exact opposite.

They build first, then treat sales as the annoying work that comes after. Many almost hate selling, so they hide behind “PLG” and live in a fantasy world where they convince themselves that a great product will somehow sell itself.

In the early days, especially when going from 0 to 1, you sell first and build later.

Selling gives the rocket fuel. Building makes the rocket lighter and faster. But if you cannot bring enough fuel to reach escape velocity, the rocket will never leave the ground. At that point, you are just building a beautiful art piece, not a company.

Before AI, the ability to build was the rare earth mineral. Today, AI is rapidly turning it into a commodity. The scarce skill is selling - to customers, employees, investors, and partners.

You do not need to be loud, polished, or extroverted. But someone on the founding team must know how to sell.

A team where no founder can sell will fail.

[Sacha Martinelle @chasnel_m]

Pitch day

[Sacha Martinelle @chasnel_m]

This is amazing, definitely lacking.

[Max Marchione @maxmarchione]

We just launched an at-home sperm analysis.

This test is covering 11 biomarkers, including DNA fragmentation, motility, morphology and concentration.

Collect at home and get the data directly.

Male fertility is dramatically under-measured and we'll fix that.

[Sacha Martinelle @chasnel_m]

My approach has been to try a lot of things early on. That’s the only way you can discover your preferences and needs. Once you spent a few years exploring your preferences and expanding your collection of experiences you can take active choices and build toward the life you really want.

[Sacha Martinelle @chasnel_m]

What’s your take on freemium for b2c SaaS? Do or don’t?

[Sacha Martinelle @chasnel_m]

@MaximMatias1 @theresidency in Vienna has one. Absolutely fantastic.

[Sacha Martinelle @chasnel_m]

@ChihYang04 No pork belly?

[Sacha Martinelle @chasnel_m]

@austinxwalker Good advice. I just started going public and it changes so much. Now starting to develop the habit of posting daily and sharing the progress.

[Sacha Martinelle @chasnel_m]

Very different economics, much slower speed of iteration, much larger barriers to adoption... The age of information is in a maturing phase when the age of physical intelligence is just starting.

[Sara Hooker @sarahookr]

My husband asked me yesterday why is AI moving at lightspeed while robotics feels slower?

I think its a really interesting question and speaks to what unlocked mainstream AI over the last decade—and why physical robots can't easily tap into those same accelerants.

[Sacha Martinelle @chasnel_m]

A fascinating aspect of developing 3D games with AI is discovering how beauty, feeling and technique are related.

You can say taste is subjective but there is a large correlation between technical proficiency and beauty.

Painters who deconstructed the rules of art like Picasso or Dalí also had very high levels of technical proficiency.

Feeling is a bit different, a piece of art can be technically impressive, beautiful, but induce a feeling that is not what you desire.

Teaching AI how to build art is taking into account all of these dimensions.

[Sacha Martinelle @chasnel_m]

@vincent_maurin @joinswanbase Premiers feedbacks utilisateurs et pleins de choses à construire ! Merci @swanbase pour tous les conseils growth 🙏

[Sacha Martinelle @chasnel_m]

@gabriel1 Incredible

[Sacha Martinelle @chasnel_m]

You can sprint on a cycle too. Not a fair comparison.

[William A. Wallace, Ph.D. @WilliamWallace]

Exercise guidance is written in minutes per week. Three minutes of sprinting moved a hundred times more blood protein than ninety minutes of cycling.

Working muscle does not only consume fuel. It secretes. Muscle, fat, liver and immune tissue release proteins and metabolites into circulation during and after exercise, collectively called exerkines, and these signals are the leading candidate for how a local activity in the legs produces effects in the brain, pancreas and blood vessels. The scale of that release is substantial. Profiling 4,163 plasma proteins in 75 middle-aged adults during a treadmill stress test found 765 of them changed at peak effort. What nobody had done was separate intensity from duration in the same people.

Manufacturing a protein and exporting it takes hours, yet the sprint response appeared immediately. The proposed route is that many of these proteins were never newly made. Pieces of proteins already anchored on cell surfaces are cut loose and released directly into circulation, a process requiring no new synthesis and therefore no delay. The team also mapped which tissues the regulated proteins likely came from and which they were heading toward, and found fat cells especially reactive: human adipocytes bathed in plasma drawn after sprinting underwent broad shifts in gene activity, including in how they handled fuel and sensed hormones.

Participants completed six all-out 30-second sprints, three minutes of genuine work, and separately performed 90 minutes of continuous moderate cycling. Of 2,884 plasma proteins measured, 714 changed immediately after the sprints, roughly one in four. Seven changed after the moderate session. Cross-referencing against a large population database, 32 of 33 proteins associated with lower metabolic disease risk were moved by sprinting, against three by moderate exercise. The response persisted after eight weeks of training.

These are small cohorts measured acutely, and a blood protein changing is not a health outcome. The disease associations come from an entirely separate population dataset layered on top, so they establish that these proteins track with lower risk in other people, not that moving them lowers anyone's risk. The comparison also confounds intensity with modality and duration at once, since sprints, cycling and running differ in more than how hard they are. Most importantly, the moderate session was not inert, it was slower: fatty acids and liver-derived proteins rose several hours afterward rather than immediately, which means a snapshot taken right after exercise is structurally biased toward whichever stimulus acts fastest. Reading the immediate window as the whole story would overstate the gap.

Twenty years ago, six sessions of sprint intervals over two weeks, totalling two and a half hours of training against ten and a half, with about 90 percent less total work performed, produced similar improvements in muscle oxidative capacity and time-trial performance to continuous endurance training. The molecular finding here sits underneath that older result rather than replacing it, and it offers a plausible reason why so little work produced so much adaptation.

What remains untested is whether a larger immediate protein response produces better long-term health, which would require a trial comparing intensities on actual outcomes rather than on blood. Nobody has run it. The closest real-world evidence comes from accelerometer data in 3,293 US adults who did no structured exercise at all, where brief vigorous bursts during ordinary daily life tracked with lower mortality in a dose-response pattern that flattened quickly.

Intensity is a lever most people are not pulling, and that it appears to be doing something chemically distinct rather than simply more of the same. That is not a case for abandoning moderate exercise, which carries the strongest outcome evidence and the lowest injury risk, and it is not a case for anyone with cardiovascular disease sprinting without medical guidance. It is a case for treating hard efforts as their own category rather than as an optional upgrade on volume.

Olsen et al., Cell Rep Med 2026;102988 · PMID 42594877
Mi et al., Mol Cell Proteomics 2023;22(8):100601 · PMID 37343698
Gibala et al., J Physiol 2006;575(Pt 3):901-11 · PMID 16825308
Koemel et al., Int J Behav Nutr Phys Act 2026;23(1) · PMID 41612409

[Sacha Martinelle @chasnel_m]

Test en live. On est là jusqu’à 18h !

[Sacha Martinelle @chasnel_m]

Mise en place de l’événement - rdv à 15h au 127 rue d’Aboukir !

[Sacha Martinelle @chasnel_m]

Mise en place de l’événement - rdv à 15h au 127 rue d’Aboukir !

[Sacha Martinelle @chasnel_m]

@Quentin_Rnv 💪🇫🇷

[Sacha Martinelle @chasnel_m]

@notaustil @JuliusCasio @joinswanbase https://t.co/lXhRVzsjaU

[Sacha Martinelle @chasnel_m]

I’m launching https://t.co/TTBUEsLo9P with @JuliusCasio to make indoor cycling 100x more immersive.

We’re at @joinswanbase and looking for beta testers to rapidly improve the product.

Come join us in the 2ème arrondissement!

[Sacha Martinelle @chasnel_m]

@notaustil @JuliusCasio @joinswanbase Je parle de demain ! On est pas au bureau aujourd’hui :)

[Sacha Martinelle @chasnel_m]

@notaustil @JuliusCasio @joinswanbase On reste toute la journée, on a mis l’event entre 15h et 18h mais il y a un peu de flexibilité