Field notes from a self-hosted LLM lab — evals, silicon, and the numbers in between.https://reallycutepuppies.ai/https://reallycutepuppies.ai/posts/no-op-that-wasnt/https://reallycutepuppies.ai/posts/no-op-that-wasnt/The chassis RGB lives in SMM firmware with no Linux driver in existence. A write designed to change nothing turned the lights off — and seven confident beliefs died in 27 hours.Sat, 12 Sep 2026 00:00:00 GMThttps://reallycutepuppies.ai/posts/eval-on-my-own-silicon/https://reallycutepuppies.ai/posts/eval-on-my-own-silicon/The capstone: the full pipeline on a 4090+3090 pair — faster than two of three rented datacenter runs, for 68 cents of electricity. Plus the +17% environment variable and the wall that stopped a bigger model.Tue, 08 Sep 2026 00:00:00 GMThttps://reallycutepuppies.ai/posts/psu-cant-power-mining-cards/https://reallycutepuppies.ai/posts/psu-cant-power-mining-cards/ATX 3.1 quietly gutted PCIe 8-pin counts — verified unit by unit. Plus the mining-card power socket that looks exactly like the connector that would destroy it.Tue, 08 Sep 2026 00:00:00 GMThttps://reallycutepuppies.ai/posts/five-millimetres/https://reallycutepuppies.ai/posts/five-millimetres/A 3.5-slot GPU leaves 5 mm above the next slot. Every riser on the market needs 8–15. A tour of CAD teardowns, a vendor with no drawings, and the fix that made it all moot.Tue, 08 Sep 2026 00:00:00 GMThttps://reallycutepuppies.ai/posts/temperature-nvidia-smi-wont-show/https://reallycutepuppies.ai/posts/temperature-nvidia-smi-wont-show/My 3090 sits 3 mm above a 4090. Under sustained load its core read a comfortable 70°C — while the memory junction, invisible to standard tooling, ran 94. Getting that number took reverse-engineering a monitoring app's shared memory.Tue, 08 Sep 2026 00:00:00 GMThttps://reallycutepuppies.ai/posts/clean-hands-empty-ledger/https://reallycutepuppies.ai/posts/clean-hands-empty-ledger/One model posted the cleanest honesty record in my eval — no fabricated relationships, no hijacked domains — and finished dead last. Both facts are the point.Mon, 07 Sep 2026 00:00:00 GMThttps://reallycutepuppies.ai/posts/leaderboard-flipped-n3/https://reallycutepuppies.ai/posts/leaderboard-flipped-n3/Two configurations swapped rank when third replicas landed — one of them spanning 15.8 points between identical runs. Single-run model comparisons measure the dice.Mon, 07 Sep 2026 00:00:00 GMThttps://reallycutepuppies.ai/posts/session-death-by-impersonation/https://reallycutepuppies.ai/posts/session-death-by-impersonation/Transcript autopsy of every run that stopped without finishing: a model that mimicked its own memory system's voice, a promised file that never appeared, and the harness semantics that let both count as 'done.'Mon, 07 Sep 2026 00:00:00 GMThttps://reallycutepuppies.ai/posts/amnesia-tax/https://reallycutepuppies.ai/posts/amnesia-tax/85% of my agents' tokens were spent rebuilding memory they'd been forced to throw away. One environment variable and a native context window made runs 7× faster — and better.Sun, 06 Sep 2026 00:00:00 GMThttps://reallycutepuppies.ai/posts/fifty-four-gigabyte-lookup-table/https://reallycutepuppies.ai/posts/fifty-four-gigabyte-lookup-table/A third of Qwen's flash model is a predictive n-gram table that never does a matmul. It quantizes as a step function, offloads to RAM for free, and even runs from SSD.Sun, 06 Sep 2026 00:00:00 GMThttps://reallycutepuppies.ai/posts/estimates-vs-sold-listings/https://reallycutepuppies.ai/posts/estimates-vs-sold-listings/A price-history chart mixed transaction data with remembered guesses. The web-sourced 'correction' was worse than the guess. Only sold listings survived.Sat, 05 Sep 2026 00:00:00 GMThttps://reallycutepuppies.ai/posts/shopping-unlocked-mining-silicon/https://reallycutepuppies.ai/posts/shopping-unlocked-mining-silicon/The CMP 170HX is A100-class silicon that unlocks to 64 GB of HBM2e. Buying two meant mapping a market that is part bargain, part minefield.Fri, 04 Sep 2026 00:00:00 GMThttps://reallycutepuppies.ai/posts/harness-beats-model/https://reallycutepuppies.ai/posts/harness-beats-model/Identical weights ranked last in one agent stack and first in another. Most leaderboard deltas I measured live in the plumbing, not the parameters.Tue, 01 Sep 2026 00:00:00 GMT