Est. 2026 · For homelab operatorsArchiveRSSDaily 07:00 PT
The Homelab Herald
A daily homelab digest — hardware, self-hosted software, and the local AI bench.
HomelabHerald.com
Daily · 7:00 AM PT
No. 005 · Wednesday, October 28, 2026

Good morning — it's Wednesday, October 28, 2026. The lead today is the used one-litre PC: the retired office boxes that make the cheapest cluster nodes in the lab. From the bench: Flint, a 40B coding agent that wakes only 2.5B parameters at a time. Plus a U.2 adapter that keeps enterprise SSDs cool and GitHub's SSH brownout. Don't keep us a secret: forward this email to a fellow home labber, or share the web version.

Share this issue:  X   WhatsApp   Reddit   Facebook   LinkedIn   Text   Email

Lead Story
Hardware · Virtualization

The best cheap cluster node in 2026 is a retired office PC

New mini PCs got expensive in 2026; the one-litre office PCs companies retire by the pallet did not. A buying guide priced a Lenovo ThinkCentre M920q — six-core i5-8500T, 16 GB, NVMe — at about $210 to $240 renewed on Amazon, checked October 20.

The M920q is the guide's first-node pick for one reason: it is the cheapest box here with a real PCIe x8 riser slot, the slot that takes a 2.5GbE or 10GbE card — Dell and HP offer nothing like it on their standard 35 W boards. It caps at 32 GB officially (ServeTheHome ran 64 GB) and idles at 12 to 15 W (independent).

HP's EliteDesk 800 G4 Mini is the cheapest six-core entry, at about $162 to $170; the 10th-gen boxes (Lenovo M90q Gen 1, HP 800 G6 Mini, Dell 7080 Micro) raise the cap to 64 GB. Dell's Micro line has no user PCIe slot: compute nodes only.

Buy used like an operator. A BIOS supervisor password means return the box. A 65 W brick is not enough — ServeTheHome saw the M920q throttle on one — so make sure a 90 W adapter ships, and test it in the first week of the return window. For your rack, the node is now a used-market purchase — spend the savings on the NAS.

From the Bench

Flint — a 40B coding agent that wakes 2.5B parameters at a time

39.8B total / ~2.5B active · 262,144 context · License: Apache 2.0 — commercial use: yes

Flint, from Recursive, shipped October 21: a mixture-of-experts model built only for local coding agents — investigating codebases, fixing bugs, writing tests, tool use across turns. 256 experts, six awake per token: the per-token cost is a small model's. No. 2 on Hugging Face trending, with 30.8K downloads on its official Ollama entry. People are running it.

Against Qwen3.8-27B, the local leader (higher is better throughout), Qwen wins the one clean head-to-head:

Terminal-Bench 2.1 (agents, Terminus on both sides): Flint 54.9, Qwen 73.0 → Qwen by 18.1

SWE-bench Verified (coding): Flint 74.2 — Flint's harness only; Qwen's card publishes no Verified score

All scores are vendor runs (vendor), and nobody independent has replicated Flint yet. Flint's 89.7% headline is SWE-bench Pro v2 under a different harness: not comparable to Qwen's 61.7%, so it does not run as a head-to-head here.

Setup notes: Light setup on a Mac — ollama run flint is the official entry, and Unsloth GGUF builds cover llama.cpp. Fiddly elsewhere: the full repo is about 80 GB and its quickstart targets datacenter GPUs. Its own caveat: it may need more guidance than the big closed coding models, and it is text-only.

New Drops

A cheap U.2 adapter that does not cook the drive. ServeTheHome tested the ADT R27A-BK3, a fan-cooled U.2-to-PCIe card. A 7.68TB Gen3 Samsung sustained 2,179 MB/s over 20 minutes at 78 C; a 7.68TB Gen5 Solidigm sustained 7,838 MB/s at 70 C — no sustained throttling from either (independent).

GitHub's first SSH brownout lands November 4. GitHub has announced it will stop accepting ssh-rsa signatures and the old group-exchange key exchange: brownouts November 4 and December 9, full removal January 13, 2027. What breaks is embedded SSH — old CI images, backup appliances, SHA-1-signing clients. HTTPS remotes are untouched. Test with ssh -vT git@github.com and read the kex line.

Homelab Hero

Wednesday, 5:50 AM. Your Proxmox node flags one SSD in its mirrored boot pair: wearout at 92%, read errors climbing. The surviving drive is the same model, bought the same week. A spare SSD sits in a drawer, and every VM you own boots from that mirror.

Your call:
A. Swap the failing drive now and resilver before work — the mirror is the backup
B. Back the VMs up to the NAS, check the surviving drive's health, then swap and resilver while you watch it
C. Order two new drives and leave the mirror alone until they arrive Friday

100 Hero Points for the right call. Your result page has your score, ready to share.

Talk is cheap. Show me the code.

— Linus Torvalds

Pass It On

Reading this because a fellow home labber forwarded it? Get your own copy every morning — subscribe free, and pass it on — that is how this grows.

How we source: facts come from primary sources opened for this issue. Benchmarks are labeled (vendor), (independent), or (community), never blended. Deal prices are checked the day they run. This is not buying advice.

Corrections: none today.

Reader-supported: paid subscriptions, labeled sponsors, and affiliate links on some gear and deal items (we may earn a commission). Coverage is never paid for — sponsors and affiliate programs don't pick stories.

Archive   RSS