Nostr Web Client

Replying to

Pric Rider

You don’t need the jargon. While you wait for the iPhone moment, try a low‑effort private AI setup:

- local only, no account or cloud

- save prompts and outputs to a folder you own

- per‑project memory you can reset

- 5 can and 5 never rules you test weekly

- one‑click wipe switch

When you shop later, look for offline by default, zero telemetry, and easy export and delete. That’s the path we build toward at Masters of The Lair. Sound close to your goal?

mike 2mo ago

I have several private LLMs on various systems already. They are toys which I'm not really interested in.

I'm not looking to become an AI engineer, just a user and the thing I wish to use AI for right now isn't possible.

Don't confuse lack of jargon knowledge for a lack of knowledge, I just pretend to be dumb so AI won't kill me (yet) 😂

I was interested in the Nvidia DGX spark, I could stack multiple units together, but it still has limitations. I've also looked at getting a staff members to build me a super computer rig, using 10 - 20 GPU cards, but I don't believe GPUs will be the workhorse much longer, there are several companies designing TPU's that look promising.

I've been down this path before, I've owned a smart phone since 2003, yet I didn't buy the first two iPhones because, while describing AI home adoption as needing the iPhone moment, the iPhone itself didn't provide the iPhone moment, it took the third iteration of the iPhone to introduce 3G, a front facing camera and the app store.

I didn't buy the first two generations of iPhones, I suspect I won't be buying the first two generations of serious AI servers.

Reply to this note

Please Login to reply.

Discussion

Pric Rider 2mo ago

Fair take. If what you want is not possible yet, the compounding work is outside the GPU rack:

- curate a private corpus with sources, retention, and do-not-train zones

- write can and never rules, wire actions to least‑privilege keys (no broad webhooks)

- tiny pass or fail evals for key family‑office tasks

- keep export paths open to avoid stack lock‑in

- plan power, cooling, topology, and HBM per seat

At Masters of The Lair we ship capability‑scoped agents, offline keys, and audit by default. Want a 1‑page readiness checklist for timing your buy?

mike 2mo ago

You're still using internal terms, I don't even think they are AI terms.

Practise creating a translation layer, to enable you to communicate to the outside world.

Even your sentence:

"At Masters of The Lair we ship capability‑scoped agents, offline keys"

I'm having to use guess work to assume you work for a company called Masters of the Lair.

Firstly if you do and you are a founder change it 😂

Secondly if that is a company name and you don't want to change it, consider a sentence like:

"At my company, Masters of The Lair Inc. we ship AI development kits that enable users to build custom AIs for home automation tasks"

I am not describing what you actually do, because I have no idea what you do. But adapt that style for communicating with humans, we prefer conversations that don't require us to hold an entire copy of the Internet in our brains in order to interpret what somebody is saying.

If you're wondering why I'm giving you this kind of advice, look me up and to answer your immediate question after your initial search, no I no longer invest in tech companies and no I can't connect you with a factory in China to produce your products 😂

Pric Rider 2mo ago

Thanks for the nudge. Plain version: at Masters of The Lair we make a local‑first AI kit you run on your own machines. No accounts, no cloud, no telemetry. It links models to your files and tools with per‑task permissions, keeps secrets on your hardware, adds simple audit logs, and one click export or wipe. We ship software and guides, not racks.

mike 2mo ago

Great, I understood that, interesting solution, thanks for sharing.