Haunted Business, Episode 7: something stirs in Bea's old laptop. π AnythingLLM, a free app, runs a small AI model right on hardware she already owns, so she can ask about her notes and docs. No monthly fee. See the setup (starter AI on your laptop) on https://t.co/eq4XaLXREZ. Follow so you don't miss the next episode. π»
#halloween #ai
Rent a 24GB GPU server from $174.50/mo instead of buying a $3K+ AI PC. Database Mart RTX A5000 24GB: from $174.50/mo at Database Mart (24-mo term, checked Oct 19). Renting is the other way in. https://t.co/dxXqa0g4r7 #ad#gpuserver#cloudgpu
Your own private AI server for $48 a month. DigitalOcean Basic Droplet (8GB RAM, 4 vCPU): $48/mo at DigitalOcean (checked Oct 21). https://t.co/Rsp2naKxWc #ad#digitalocean#selfhosted
@gpbulkin Heads up if you're on NVIDIA: Ollama's Flint build (v0.40.8) runs on Macs only for now, with other GPUs coming later. It's a 23GB download, so plan on a 24GB card or llama.cpp with a few layers offloaded to RAM.
Is your AI box password on a sticky note? Your private AI box has a pile of admin logins: OS, router, Open WebUI, NAS, backups. Give each one a long unique password and turn on 2FA. https://t.co/0B3aK4qE8V #ad#passwordmanager#cybersecurity
Is the RTX 5060 Ti 16GB enough for local AI? Find your level. ASUS PRIME RTX 5060 Ti 16GB: $869.99 at ASUS eShop (checked Oct 19). https://t.co/2dhtaiGzuU #ad#rtx5060ti#nvidia
@RizRizawan Used RTX 3090s are still the budget way to get 24GB of VRAM. New 16GB NVIDIA cards are pricey right now. If that stings, rent a cloud GPU by the hour first, then buy once you know which models you actually use.
Haunted Business, Episode 6: Bea's thumbs give out after hours of typing. An office PC turns her voice into text on iPhone, iPad, Android and desktop in about 2 seconds, with unlimited use and nothing leaving her network. See the setup on https://t.co/eq4XaLXREZ π» #halloween #smallbusiness
π¨ URGENT: Claude can now CREATE a YouTube channel from scratch and take it to monetization in just 90 days. 100% free.
Here are 8 prompts to achieve it:
Does your AI box have a backup? MiniTool ShadowMaker Pro Ultimate: $79 one-time (checked Oct 19). One dead drive can wipe out your whole private AI setup. Windows 11/10/8.1/8/7 image and file backup. Windows only. https://t.co/0B3aK4qE8V #ad#backup#windows
@IArbilly Yes, ship it. Plenty of people want agents that never touch the cloud. Make first run painless: detect Ollama or LM Studio, suggest a model that fits their RAM, and say clearly that nothing leaves the machine.
@jetpen@cometaj2 Bitwarden, 1Password and Keeper all sync across laptops and phones with apps on every platform. If you share logins with family or a team, Keeper's shared folders work well, and Bitwarden's free tier is a solid way to start.
@ayubhkd Full DeepSeek V3 is 671B. In Hardware Corner's test, 4-bit peaked at ~466GB on a 512GB M3 Ultra Mac Studio and dropped to ~6 tok/s at 16K context. Breakdown: https://t.co/nmkn549msn
What my shop pays for AI vs. paying once. Mac mini M6 (16GB, 256GB): $779.99 at Micro Center (checked Oct 19). https://t.co/z2uTKizGlT #ad#macmini#applesilico
@popularaiorg For local LLMs, VRAM usually wins. 16GB fits 14B models at Q4 with room for context, while 12GB gets tight fast. The 5070 is quicker on what fits, but a 5060 Ti 16GB like the ASUS PRIME loads more models.
@hitch_hiker_dev pass is the classic: GPG-encrypted files, git sync, nothing in anyone's cloud. Bitwarden (bw) and 1Password (op) both have solid CLIs too. If your team's on Keeper, Keeper Commander gives you a full CLI plus scripting.
@TSC Depends what you want to run. A 16GB Mac mini (the M6 is $779.99 at Micro Center right now) handles 7-8B models fine. For 30B-class models you want 48GB+ unified memory, like the M4 Pro 48GB.
Haunted Business EP 5: product descriptions are due and the page stays blank. A free AI on Bea's own computer rewrites, shortens and fixes her draft in LibreOffice, and nothing goes to the cloud. π»
Setup: https://t.co/CsIU4xRPg7
#halloween#smallbusiness