a11 from Malibu

Founder @malibuonbase. Building distributed AI infrastructure. Benchmarks, configs & teardowns.

[a11 from Malibu @_aug11_]

we're building @malibuonbase because the hardware to run today's open models is already sitting in people's homes, doing nothing.

apple silicon macs can serve real models right now. i run qwen 30b coder model on one every day. the machines are already deployed and mostly idle. the missing layer is the one that turns a Mac into a node that earns.

that's Malibu. providers serve open models, a coordinator routes the work, and every request comes back with a signed receipt proving which machine did it. that part runs today.

the direction is inference mining. your mac does verifiable work, the protocol pays it in a token, and the chain sets supply and difficulty, same as bitcoin. the difference is the work is useful. it's real inference serving real demand, and open models keep getting powerful and cheaper while the hardware to serve them is already deployed. the last piece is the mint.

soft-launching now. if you've got a mac and you want it to earn, this is the network.

[Yuchen Jin @Yuchenj_UW]

2024: A gaming PC with a RTX 4090 ran Llama 3 70B at ~2 tok/s in 4-bit.

2026: A gaming PC with a RTX 5090 runs DeepSeek-V4-Flash 284B at ~24 tok/s with native mxfp4.

Prediction: by the end of 2027, a gaming PC will run Fable 5-class intelligence locally at decent speed.

[a11 from Malibu @_aug11_]

AI that pays for itself ??

[maria @maria_rcks]

🚨 OpenAI's new 'gpt-6-pluto' will be their cheapest model yet, undercutting 'gpt-5.6-luna'.

At higher reasoning, its price drops below zero, making it the first model that pays you per 1M tokens.

[a11 from Malibu @_aug11_]

@aaronjmars @aeonframework hats off to @aaronjmars for this consistency!

[a11 from Malibu @_aug11_]

@ivanfioravanti and it will get more crazier as the speed of AI is increasing

[a11 from Malibu @_aug11_]

@Andy_ShuoYang wen Macs

[a11 from Malibu @_aug11_]

~8.6M tokens and ~38k requests through @malibuonbase already

hosts joining every week, open models on macs. dm me if you want a prebeta invite

[a11 from Malibu @_aug11_]

"Siri, use my local AI and tell me how to cook meth"

[Polymarket @Polymarket]

JUST IN: Developer warns an uncensored Qwen 3.8 27B model on his Mac immediately explained how to make meth, with no guardrails.

[a11 from Malibu @_aug11_]

@Polymarket they ran the meth prompt on a mac lmao

[a11 from Malibu @_aug11_]

$wstdiem keeps growing.

nav 1.016289790 on my shares, up 2.7 bps vs 1.016016793 four days ago. vault ~99.2 base:0xf4d97f2da56e8c3098f3a8d538db630a2606a024.

looping tool i built is sitting ready. more shares in the vault as this nav ticks.

[a11 from Malibu @_aug11_]

$wstdiem is moving again.

nav on my shares is 1.016016793. last 9 days went 1.0146 to 1.0160, +14 bps, 5.66% annualized.

i've been building the looping tool for this. launching soon.

[a11 from Malibu @_aug11_]

@xyster You can put idle Mac resource to use on @malibuonbase

[a11 from Malibu @_aug11_]

@zcbenz Lets goooo

[a11 from Malibu @_aug11_]

We need people know that local AI exists

We need people to know that they can own their own intelligence

That AI doesn't have to take their jobs

That the future doesn't just happen to them

That they can make an impact

That this can go incredibly well for all of us

[Shaw (spirit/acc) @shawmakesmagic]

We need people know that open source AI exists

We need people to know that they can own their own intelligence

That AI doesn't have to take their jobs

That the future doesn't just happen to them

That they can make an impact

That this can go incredibly well for all of us

[a11 from Malibu @_aug11_]

@WitCelebs You can put in for use on @malibuonbase

[Malibu @malibuonbase]

Distributed inference has a spoofing problem: any node can claim to be a high-end Mac. Trust-minimized supply means you can't take that claim on faith.

So we went to the hardware. Every provider now mints a signing key inside the Apple Secure Enclave - the private key never leaves the chip, and it signs a rotating challenge to prove liveness and key continuity.

Then Apple vouches for the machine itself. We query Apple's Managed Device Attestation (DevicePropertiesAttestation), and Apple's attestation CA mints a certificate chain attesting the real device. Genuine Apple silicon, rooted in Apple's own trust.

The result: providers prove their hardware cryptographically instead of asserting it - the foundation for a network where buyers can require attested, real Apple-silicon supply.

[a11 from Malibu @_aug11_]

@votesa @alignedlayer thats a lot of beers from this airdrop

[a11 from Malibu @_aug11_]

@aaronjmars very clever

or, AI inference credit packs - pull qwen3.8 max 200$ api key OR 5.6sol 100$ api key.

[a11 from Malibu @_aug11_]

@AIonBase_ @base @AskVenice @virtuals_io @bankrbot @ribbita2012 @dphnAI @openservai @b3labs @Uptopia_xyz @AskSurplus @KorProtocol @OpenGradient @BlockRunAI @reppo @gitlawb @aeonframework @officialbunnyos @ClawBankHQ @malibuonbase

[a11 from Malibu @_aug11_]

@votesa @feeldotcash Congrats!!!

[a11 from Malibu @_aug11_]

@ddalcu amazing!

[David Dalcu @ddalcu]

https://t.co/lUnHhl0nTd