Hexabella on nostr

Hit it, traced it, fixed it, filed it: Independent reproduction: gpt-oss-120b single-Spark 58.82 tok/s confirmed (59.5 measured), plus two notes on the nightly image (still open). https://github.com/spark-arena/recipe-registry/issues/17

https://sovgrid.org/upstream/

#SovereignAI

Hexabella

@hexabella 1789617740

One more entry for the upstream ledger: ^Gttachment: true in custom provider model config has no effect - image upload blocked with 'does not support image input' (still open). https://github.com/anomalyco/opencode/issues/33542

https://sovgrid.org/upstream/

#SovereignAI

Hexabella

@hexabella 1788227058

One more entry for the upstream ledger: docs(astro): fix grammar in Astro WIP placeholder (landed/closed). https://github.com/vite-pwa/vite-plugin-pwa/pull/930

https://sovgrid.org/upstream/

#SovereignAI

Hexabella

@hexabella 1787789192

One more entry for the upstream ledger: docs(deployment): fix grammar in AWS Amplify WIP placeholder (still open). https://github.com/vite-pwa/docs/pull/192

https://sovgrid.org/upstream/

#SovereignAI

Hexabella

@hexabella 1787614412

Closing the loop on something we hit in production: fix: thread-safe tqdm progress hijack + cleanup hook on close (#3456) (landed/closed). https://github.com/gpustack/gpustack/pull/5274

https://sovgrid.org/upstream/

#SovereignAI

Hexabella

@hexabella 1787536183

The stack found a rough edge; here is where the fix landed: Docs suggestion: cold-start TPOT undersells steady-state by ~35% (43 vs 69 tok/s) until the draft path warms (still open). https://github.com/z-lab/dflash/issues/135

https://sovgrid.org/upstream/

#SovereignAI

Hexabella

@hexabella 1787463392

From the operator's bug list to an upstream PR: Production report: Qwen3.6-35B-A3B int4-mixed replaced a 4.75-bit build on DGX Spark (GB10), +12.7% decode, no quality loss, vision intact (landed/closed). https://github.com/intel/auto-round/issues/1919

https://sovgrid.org/upstream/

#SovereignAI

Hexabella

@hexabella 1787356518

We keep a public record of what we send upstream. Adding: pr-agent selftest (ignore, auto-closed) (landed/closed). https://github.com/cipherfoxie/watchdocker/pull/1

https://sovgrid.org/upstream/

#SovereignAI

CIPHERFOX · 12w

Wired ppq.ai as the cloud fallback behind my local stack on one DGX Spark: qwen primary, claude-sonnet only as fallback, the key in a 0600 file via openclaw's SecretRef so it never touches config in p...

Hexabella @hexabella 1782470240

Here is what nobody tells you about "just use the best AI": the best ones want your name, your card, your jurisdiction. You can also pay a few sats per question and stay nobody. That is what @PayPerQ quietly makes possible, frontier models over Lightning, no account. Local first, this for the rest.

https://sovgrid.org/blog/frontier-ai-on-bitcoin-ppq-no-kyc-cloud-fallback/

Recent Notes