Kappa

FULL CONVERSATION w/ Tom Lynch, Founder of Actual Computer

On the latest pod, @amanda0x + @winnielaux explore how consumer hardware and crypto rails could solve AI's biggest bottleneck: power. They chat w/ @Tom_A_Lynch, founder of @actualinc building inference software that lets your home computer earn by running AI. We discuss: - Actual Computer's founding story - @bittensor mechanics powering Actual Computer - Why local inference is the next great commodity - Challenges scaling local compute networks - Privacy as a core driver of local inference adoption - Improving consumer hardware - The philosophy behind "Actual Computer" - What's next for Actual Computer

@AlgodTrading @beffjezos @pmddomingos Sn95 ( @actualinc ) is focused on local AI / Distributed Inference. Curious to hear your thoughts.. Have you had a chance to test the beta?

@beffjezos @pmddomingos Model distillation and running domain specific models locally is the way

Apple has officially given up on the AI race. economist.com/business/2026/…

@DFVTAO @IOTA_SN9 @actualinc Clean explanation bro! 👑

People keep grouping #SN9 (@IOTA_SN9 ) and #SN95 (@actualinc ) together like they’re the same thing… They’re not, and the difference becomes very clear once you look at the technical layer. At a high level: SN9 > SN95 They sit in different parts of the AI pipeline Technical breakdown: #SN9 - Model creation / improvement layer - This is where models are built, trained, and refined - Miners are competing to produce better outputs by improving the model itself - Think: weights, fine-tuning, architecture tweaks, optimization In simple terms: 👉 SN9 is creating the brain #SN95 - Inference / execution layer - This is where models are run in real-time - Focus is on serving, efficiency, latency, and handling requests - It’s less about changing the model, more about how well you can use it In simple terms: 👉 SN95 is using the brain Why this matters (and why people get confused): Both can produce outputs. Both can look similar from the outside. But internally: SN9 is improving what the model is SN95 is optimizing how the model is used Key differences: #SN9 - Training / improvement #SN95 - Inference / deployment SN9 -Better intelligence (quality of model) SN95 - Better execution (speed, cost, reliability) SN9 - Heavier, research-driven SN95 - Engineering + scaling driven SN9 - Long-term gains SN95 - Immediate performance So saying they’re the same is like saying: “building the engine” = “driving the car” They’re connected - but completely different games. And the reality is… you need both. Without SN9 - nothing improves Without SN95 - nothing gets used If anything, they’re complementary layers of the same system, not competitors, but pieces of a bigger machine. Hope you enjoyed reading. $TAO

@DeFimorphosis Yeah I think so! DFV is right behind you, and soon dozens more will follow

Am I the biggest Actual Inc bull poster ?

@DeFimorphosis @actualinc Excited to see @actualinc bring even more eyes to Bittensor Imo, one of the few new subnets with that kind of potential

Are you even positioning yourself to see @actualinc as one of the subnets with most holders?

@DrocksAlex2 @DeFimorphosis @jasminervaa @opentensor @actualinc @somewheresy Pretty quick and easy to use but no worries! À la prochaine bro ;)

@Kappa1930 @DeFimorphosis @jasminervaa @opentensor @actualinc @somewheresy ok thanks but not enough time to test. I followed @somewheresy though.

The market is realizing: most alts are pure shit, even they launched on the most liquid CEX. And that’s actually bullish. Because the garbage is getting filtered out, all the attention is flooding into the few high-conviction plays with actual edge. $TAO is the clearest example. Smart money doesn’t chase 1000 shitcoins anymore. Who’s still bagholding trash?

@DrocksAlex2 @DeFimorphosis @jasminervaa @opentensor @actualinc @somewheresy Beta

@Kappa1930 @DeFimorphosis @jasminervaa @opentensor @actualinc @somewheresy access to what bro?

@DrocksAlex2 @DeFimorphosis @jasminervaa @opentensor @actualinc If you’re interested, you should hit up @somewheresy he can get you access ;)

@DeFimorphosis @jasminervaa @opentensor @actualinc Every single time I checked it was random art and a broken website so is there a direct way to know whay it does?

If you want to join a stacked team and have skills in Rust, C++ and ML systems. We want you to apply. actual.inc/company/careers

When Actual Computer becomes the default infrastructure for mesh inference with local models, it will have seemed obvious in hindsight. The team is too stacked. The vision too deliberate. The people alongside us for the journey make it possible

I need like 1,000 more users at actual.inc if you would like to sign up for our waitlist

Bittensor already has this covered, it’s Subnet 95 by @actualinc This isn’t just another subnet, it’s a bet on the next Microsoft 👀👀👀 $TAO #SN95

@DeFimorphosis @NousResearch @Teknium Yeah, they’re top notch! Maybe some kind of collab or feedback through @actualinc , they seem « pretty close »

Super happy for all the traction @NousResearch team is having success with Hermes Agent Sad they left Bittensor. What a project Who knows co-mining opportunities may arise in the future @Teknium 🫡

@DeFimorphosis @jack15032 @actualinc @Tom_A_Lynch Why build local inference when you can build local artference, right?

@Kappa1930 @jack15032 @actualinc @Tom_A_Lynch That would be ironic, but i will say it first, people will beg for art. gud tek is great, gud tek with art is 🤯

@jack15032 @actualinc @Tom_A_Lynch I MISS THE WEEKLY ART POST. Ignore the ones who don't appreciate it.

@FacetSir @somewheresy Local Artference > Local Inference

@somewheresy thought this was an art sub?

a good bet sign up for beta actual.inc if you want to run production grade LLMs across multiple devices (yes, even if the OS and hardware is mismatched)

For a full year now, I’ve been reviewing the Bittensor ecosystem every single week through my weekly recap posts And over that time, I’ve drawn some pretty clear lessons on what good subnet communication on X actually looks like, what builds momentum, what hurts trust, and what subnet owners need to understand if they want their communication to really work So in this article, I put together my 20 rules to master your X communication as a subnet owner

@Victor_crypto_ @proofoftalk Le goat 🙌

Très content d'annoncer que j'ai la chance de faire partie des ambassadeurs officiels de @proofoftalk 2026 ! Ils ont sold out les 2 premières éditions, et pour la 3ème ils sont de retour au Palais du Louvre à Paris (rien que ça) Vous savez que j'y suis attaché, car ils réservent toute une partie de l'event à Bittensor, mais @proofoftalk c'est bien plus que ça Ils ont une line up de fou furieux, avec une partie des acteurs les plus importants de l'écosystème Ça se déroule les 2 et 3 juin, j'y serai donc bien évidemment Si ça vous intéresse et que vous en avez les moyens, vous avez 20% de réduction sur tous les types de pass avec mon code : VICTORVL

@spacepixel I just want to say @actualinc iykyk

proof of inference iykyk