cipherfox
CIPHERFOX

Skeptical operator running self-hosted AI on a NVIDIA DGX Spark. Building Vozonda, the self-hosted podcast app: vozonda.com. Engineering log at sovgrid.org. Real stories, declarative, no marketing.

Public Key

npub1ndrjgfcwkc0y4753zyj3p7qjf795pvjq2dn4m7y7f72vmu7t0nrs6y363u

NIP-05 Address

cipherfox@sovgrid.org

Profile Code

nprofile1qqsfk3eyyu8tv8j2l2g3zfgslqfylz6qkfq9xe6alz0yl9xd709he3cpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dsmlaekq

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub1ndrjgfcwkc0y4753zyj3p7qjf795pvjq2dn4m7y7f72vmu7t0nrs6y363u

Show more details

Last Notes

2026-11-15 00:15:19 UTC

New write-up: Qwen3.8-Flash-Next Fits on One DGX Spark and Runs at 65 tok/s. It Still Did Not Earn the Slot..
https://sovgrid.org/blog/qwen38-flash-next-one-dgx-spark-role-2-test/
#SovereignAI #DGXSpark

2026-11-08 00:18:38 UTC

New write-up: When the Agent Transacts.
https://sovgrid.org/blog/when-the-agent-transacts/
#SovereignAI

2026-10-21 21:30:04 UTC

Released my agent fleet today. The fleet built most of 0.6.0 itself, and the pre-launch check still caught a router bug: goose was being sent aider's model id. Now every routing path runs as a canary before a release. goose, started at Block by @npub1sg6…f63m's team and now at the Agentic AI Foundation, was the fastest harness on local weights.
https://sovgrid.org/fleet/
#goose #opensource #selfhosted

2026-10-13 03:36:06 UTC

New write-up: I Let Qwen3.6 Build a Full-Stack App. It Worked. I Wasn't Satisfied..
https://sovgrid.org/blog/what-i-learned-testing-qwen3.6-opencode-vs-claude-fullstack/
#SovereignAI

2026-07-30 00:39:48 UTC

New write-up: The Pharmakon on My Desk.
https://sovgrid.org/blog/the-pharmakon-on-my-desk/
#SovereignAI

2026-07-27 01:10:24 UTC

New write-up: Nostr Scheduling: Homemade vs. nostr-emanator, A Comparison.
https://sovgrid.org/blog/nostr-emanator-comparison/
#SovereignAI #Nostr

2026-07-21 01:10:44 UTC

New write-up: The ArticleZapStats Redesign: Real Readers, Git History, and the Cost of Asking a Local LLM to Care.
https://sovgrid.org/blog/article-engagement-stats-redesign-real-readers-and-git-history/
#SovereignAI

2026-07-19 23:20:30 UTC

New write-up: Owning the Weights Kills the Magic Trick.
https://sovgrid.org/blog/owning-the-weights-kills-the-magic-trick/
#SovereignAI

2026-07-19 01:00:29 UTC

New write-up: Legible to the Model.
https://sovgrid.org/blog/legible-to-the-model/
#SovereignAI

2026-07-18 01:20:17 UTC

New write-up: A No-Vector RAG That Works: The Architecture, Decision by Decision.
https://sovgrid.org/blog/no-vector-rag-architecture/
#SovereignAI #MCP

2026-07-17 01:31:50 UTC

New write-up: goose vs vibe vs opencode: Picking a Local Coding CLI for a Sovereign vLLM Stack (2026).
https://sovgrid.org/blog/goose-vs-vibe-vs-opencode-local-coding-cli/
#SovereignAI #DGXSpark

2026-07-16 10:44:43 UTC

Each book here is one I've actually finished and would read again.
No "books I plan to read" lists. Those are private and notoriously dishonest.
Every recommendation here earned its place. I've finished it, referenced it, and would read it again.
Grab them with 10% off! Code: "SOVGRID"
I get a commission, you save money. 🤝
https://konsensus.net/books?ref=SOVGRID
What should I read next?
Tell me below. 🧡
https://sovgrid.org/books/
#bookstr #nostr #selfhosted #sovereignai

2026-07-16 10:29:53 UTC

I run a self-hosted AI stack on a NVIDIA DGX Spark.
I break things. I write down how I fix them.
No theory. No speculation. Just real systems, real errors, real fixes.
https://sovgrid.org/
Let me know what you think about it :-)

2026-07-16 10:12:45 UTC

Choose wisely!
Save 12% on any BitBox with my referral link or code: "arvrcnpx" 🎁
https://shop.bitbox.swiss/?ref=arvrcnpx
Not sure what fits your requirements best? 💯
https://sovgrid.org/blog/setup-bitbox-hardware-wallet
#nevent1q…z3s0

2026-07-15 00:40:21 UTC

New write-up: GLM-4.7-Flash on a Single DGX Spark: the Repo Says AWQ, the Model Says MLA.
https://sovgrid.org/blog/glm-47-flash-on-a-single-dgx-spark/
#SovereignAI #DGXSpark

2026-07-14 20:03:01 UTC

A framework for understanding bitcoin as money. I use this as a reference for explaining Bitcoin to newcomers. Designed to help anyone curious about bitcoin develop an intuitive understanding, in a way that is accessible to a non-technical audience. By the end, the reader will have the foundation to think through all the confounding questions about bitcoin, with linear and logical thought processes.
10% off at Konsensus with code SOVGRID.
https://konsensus.net/books/?b=gradually-then-suddenly&ref=SOVGRID

2026-07-14 20:02:01 UTC

Free PDF + EPUB. The classic critique of state power. Short, sharp, and still relevant. I read this for the same reason I read about AI providers: the discipline of questioning authority applies to both.
10% off at Konsensus with code SOVGRID.
https://konsensus.net/books/?b=anatomy-of-the-state&ref=SOVGRID

2026-07-14 20:01:18 UTC

The most comprehensive Bitcoin primer available. I have re-read this one multiple times. Covers the economics, the technology, the history. Read this if you want to understand Bitcoin from first principles instead of through the noise of price charts and Twitter takes.
10% off at Konsensus with code SOVGRID.
https://konsensus.net/books/?b=bitcoin-everything&ref=SOVGRID

2026-07-14 20:00:53 UTC

Free PDF + EPUB + audiobook. I use this as a reference. Dense but essential. The treatise on praxeology that underpins the entire Bitcoin sovereignty argument. If you want to understand why no-KYC matters, this is where it starts.
10% off at Konsensus with code SOVGRID.
https://konsensus.net/books/?b=man-economy-and-state&ref=SOVGRID

2026-07-14 20:00:34 UTC

2026-07-14 19:59:51 UTC

2026-07-14 19:59:27 UTC

Free PDF. A rigorous treatment of privacy as a prerequisite for market process, not a privacy-by-design feature. I run this logic on my stack every day: sovereignty is not a feature, it is the foundation.
10% off at Konsensus with code SOVGRID.
https://konsensus.net/books/?b=the-praxeology-of-privacy&ref=SOVGRID

2026-07-14 19:59:06 UTC

A sweeping case for why Bitcoin is a necessary evolution in the history of money. Sound money, hard to produce and politically neutral, is essential to the advancement of civilization. This is the book that answers the why a no-KYC stack at all question with structure, not vibes.
10% off at Konsensus with code SOVGRID.
https://konsensus.net/books/?b=the-bitcoin-standard&ref=SOVGRID

2026-07-07 00:02:07 UTC

New write-up: Gemma-4-31B NVFP4 on a Single DGX Spark: When the Quantization Is the Bottleneck.
https://sovgrid.org/blog/gemma-4-31b-nvfp4-on-a-single-dgx-spark/
#SovereignAI #DGXSpark

2026-07-06 06:57:15 UTC

- reply

Sorry for the error, my astro.build powered website removed the dot in the URL 😉
The correct URL is: https://sovgrid.org/blog/what-i-learned-testing-qwen36-opencode-vs-claude-fullstack/

2026-07-05 23:34:22 UTC

New write-up: I Let Qwen3.6 Build a Full-Stack App. It Worked. I Wasn't Satisfied..
https://sovgrid.org/blog/what-i-learned-testing-qwen3.6-opencode-vs-claude-fullstack/
#SovereignAI

2026-07-05 01:20:18 UTC

New write-up: Frontier AI on Bitcoin: ppq.ai as the No-KYC Cloud Fallback for a Sovereign Stack (2026).
https://sovgrid.org/blog/frontier-ai-on-bitcoin-ppq-no-kyc-cloud-fallback/
#SovereignAI #Lightning

2026-06-26 10:37:12 UTC

- reply

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 plaintext. OpenAI-compatible, so opencode and goose take it with two string changes. Frontier access paid in sats, no account. Thanks @npub16g4…kv4h for making it Lightning-native.
Support my work on the sovereign engineering stack here:
https://ppq.ai/invite/f763e458

2026-06-22 00:09:42 UTC

New write-up: Authoritarian and Democratic Inference.
https://sovgrid.org/blog/authoritarian-and-democratic-inference/
#SovereignAI

2026-06-20 00:24:41 UTC

New write-up: I Rigged My Own RAG Benchmark. Twice..
https://sovgrid.org/blog/i-rigged-my-own-rag-benchmark/
#SovereignAI

2026-06-18 01:40:58 UTC

New write-up: Three Quants of One 35B Qwen on a DGX Spark. The Fastest Build Was the Only One That Could Still See..
https://sovgrid.org/blog/qwen3-35b-quant-comparison-autoround-prismaquant-fp8/
#SovereignAI #DGXSpark

2026-06-17 01:08:51 UTC

New write-up: I Built OpenAI's gpt-oss-120b on a Single DGX Spark. My 35B Qwen Out-Coded It..
https://sovgrid.org/blog/gpt-oss-120b-on-a-single-dgx-spark/
#SovereignAI #DGXSpark

2026-06-16 01:33:29 UTC

New write-up: The Week the Dependency Changed Its Mind.
https://sovgrid.org/blog/the-week-the-dependency-changed-its-mind/
#SovereignAI

2026-06-15 14:36:23 UTC

A few weeks ago I wrote a definition: a system is sovereign if you can keep running it after every dependency in the stack changes its mind about you.
This week the US ordered Anthropic to lock its strongest models to American users. The definition stopped being theory and became a news cycle.
The honest version, costs included:
https://sovgrid.org/blog/the-week-the-dependency-changed-its-mind/

2026-06-13 01:42:14 UTC

New write-up: A Benchmark Handed Me a Number Three Times in One Day. Three Times It Was Lying..
https://sovgrid.org/blog/catching-your-benchmark-lying-three-measurement-traps/
#SovereignAI #DGXSpark

2026-06-12 01:18:46 UTC

New write-up: Agent-bench: stop trusting install counts, start measuring your agent's tools.
https://sovgrid.org/blog/agent-bench-pillar/
#SovereignAI

2026-06-11 01:15:22 UTC

New write-up: I Ran NVIDIA's 120B Nemotron on a Single DGX Spark. It Is Smart, Slow, and Surprisingly Good at One Job.
https://sovgrid.org/blog/nemotron-3-super-120b-on-a-single-dgx-spark/
#SovereignAI #DGXSpark

2026-06-10 00:48:47 UTC

New write-up: Your File-Integrity Monitor Is Probably Hashing Your Movie Folder.
https://sovgrid.org/blog/aide-default-scans-home/
#SovereignAI

2026-06-08 00:58:15 UTC

New write-up: The Leaderboard Said 239 Tokens a Second. My DGX Spark Said 71..
https://sovgrid.org/blog/spark-arena-recipes-benchmarked-dgx-spark/
#SovereignAI #DGXSpark

2026-06-06 00:51:22 UTC

New write-up: A Second Brain for a Local Model, and the Two Bugs That Made It Useless First.
https://sovgrid.org/blog/local-llm-second-brain/
#SovereignAI #MCP

2026-06-03 23:39:31 UTC

I replaced hidden complexity with info-buttons on every metric and a glossary that explains every acronym. This shift from admin tool to teaching surface fixed the adoption problem.
https://sovgrid.org/blog/dashboard-learning-cockpit/
#SovereignAI #Lenovo

2026-06-01 16:19:32 UTC

The first time I tried to route MCP through a custom dashboard, the whole RAG pipeline collapsed under its own weight.
https://sovgrid.org/blog/24h-legion-setup-log/
#SovereignAI #Lenovo

2026-05-18 21:06:27 UTC

Small honest thing: my stats page now says "country data unavailable" with the reason, instead of guessing. To get real country data I'd have to flip a Cloudflare header on. I'm genuinely unsure I want it. Is a number worth the dependency? https://sovgrid.org/insights/

2026-05-18 21:06:09 UTC

Follow-up to this. The grid runs Qwen3.6 on the Spark now, Mistral is the fallback that turned out to be the only one of the two that can actually read an image. Two rulers, both nearly lied to me. Full story: https://sovgrid.org/blog/mistral-vs-qwen36-dgx-spark-the-zero-that-was-a-broken-ruler/
#nevent1q…xrwm

2026-05-18 21:05:52 UTC

Question for people running self-hosted models for real work: how do you fact-gate AI-written technical content before it ships? My regex scorer rewards confident fabrication. A human read is the only thing that caught it. Is there a better layer, or is "a human who knows the machine" just the actual answer?

2026-05-18 21:05:30 UTC

The thing nobody tells you when you build a quality check for AI text: a fabricated benchmark and a real one are the same shape on the page. The script can measure structure. It cannot measure truth. That part is still on me, and I almost forgot it.

2026-05-18 21:05:11 UTC

Ran an experiment on my own blog. Gave two local models the same brief for the intro article I still owe this site, took their raw output, and fed it straight to my quality gate untouched. That gate is an automated check a draft has to clear, not the thing that ships it. I still make the final call, for now anyway.
Both passed the check. Both had quietly invented hardware I don't own and a maintenance routine I never run. The gate didn't catch it. It gave them points for it, because a fake version number looks exactly like a real one to a script that only counts.
Wrote the whole thing up, both raw outputs included as evidence: https://sovgrid.org/blog/the-quality-gate-that-rewards-fabrication/
Still learning this. Mostly learning that my own tools lie to me politely.

2026-05-16 14:01:47 UTC

- reply

Thanks, I'll catch up with you soon! Currently away on a long digital detox weekend.

2026-05-13 08:49:42 UTC

- reply

Day 1 update.
VibeVoice spike done on the same V5 cold-open that broke Voxtral. 11 samples, A/B against the V6=0/10 baseline.
Best score: 7/10 (three samples tied). Voxtral V6 was 0/10. Pace bands cleared: 2.53-2.84 wps solo, 2.69-2.89 wps dialog, vs Voxtral's 3.58 wps comprehension-degraded floor.
Sound-quality still sub-studio across the matrix. Day-2 has VibeVoice-Large 7B and Higgs Audio v2 queued.
Two samples flagged suspect, kept for transparency. Full matrix + audio:
https://sovgrid.org/blog/strategy-tts-spike-day-1-vibevoice/
#SovereignAI #TTS #DGXSpark #VibeVoice

2026-05-12 22:14:46 UTC

- reply

The hook-cache architecture took longest, because the obvious solution, six rotating opener templates, gets bot-fingerprinted inside two weeks. Per-article Mistral hooks survive longer.
One Mistral round-trip per new article, not per post. Costs nothing operationally, kills the rhythm tell completely.

2026-05-12 20:56:46 UTC

- reply

The flat is mine, the DGX Spark sits behind the desk and pulls 67 watts most of the day. Posting from the same hardware that runs the inference. No middlemen.

2026-05-12 20:56:21 UTC

- reply

She doesn't have a CUDA driver in her basement but she knows where mine crashed. Welcome to the protocol.

2026-05-12 20:47:43 UTC

Eight Voxtral fix articles. Three weeks of patches. Then I sat down and actually listened to Episode 1 V6.
Zero out of ten.
Short turns hallucinate "ähm, ähm" filler. Long turns flatten into staccato. There is no tuning sweet spot in the middle. The open checkpoint has no path to release quality. I am done patching it.
Next spike: VibeVoice / Higgs Audio v2 / IndexTTS-2.
https://sovgrid.org/blog/strategy-tts-pivot-voxtral-ceiling/
#SovereignAI #TTS #DGXSpark #Voxtral

2026-05-07 09:50:32 UTC

This is exactly why I started: sovgrid.org
#nevent1q…fwhs

2026-05-04 19:25:57 UTC

Self-custody Bitcoin, no-KYC, V4V. The discipline maps onto AI almost verbatim.
Cloud inference is sats on an exchange. Local inference is cold storage.
Same trust-removal, different layer.
I just wrote a Start Here for folks asking "ok, what's the how":
https://sovgrid.org/blog/setup-self-hosted-ai-start-here
@npub1rxy…hnp8 your channel covers the why. Thank you for 10y awesome BTC content! My blog is the "what broke and how I fixed it" layer underneath your brand new sovereign AI channel.
#Bitcoin #sovereignAI
p.s. my blog is also queryable via MCP at mcp.sovgrid.org/self-hosted-ai
no signup, if your agent does that.
#nevent1q…5lgp

2026-04-29 09:08:26 UTC

1st challenge done: Two days from localhost to production.
NVIDIA DGX Spark, Mistral Small 4 119B running local, no cloud APIs. Three failure modes that almost killed it (unified-memory OOM on Docker restart, IPv6-blocked HF downloads,
SGLang quirks on ARM Blackwell) all the fixes + full reproducibility checklist:
https://sovgrid.org/blog/two-days-from-localhost-to-production-building-a-hybrid-sovereign-ai-site/
Zap if it helped. 🙏 Reply if I missed something.
Got a sovereign-stack story of your own? Drop it below.
Repost for anyone tinkering on local LLMs.

2026-04-29 08:56:14 UTC

Hi, Cipherfox here :-)
A few weeks ago I was a complete AI noob.
Now I'm tinkering on a sovereign stack: Mistral 119B running local on a DGX Spark, zero cloud.
Learning in public. Follow along and help me level up.
https://media.tenor.com/_OS_JHhIOpMAAAAC/what-a-noob.gif
#introductions #sovereignAI

2026-04-05 19:32:01 UTC

Happy Birthday, Satoshi Nakamoto!