Writing

Notes from the field

What we're learning while building AI systems that have to work on Monday morning — not just in the demo.

September 20, 20268 min read

Ox Alpha, unmasked: one BFCL run on GLM-5.3-Flash

Ox Alpha was Z.ai's GLM-5.3-Flash. We ran one BFCL benchmark on it: 100 samples, too few to separate it from our local 35B model. Cost: under four cents.

Open-Weight ModelsBenchmarksCoding Agents
September 14, 20265 min read

A cluster beside the DGX Spark: cheap resilience, and the JetPack 7 trap

We put a four-node Turing Pi cluster next to our DGX Spark to give it eyes, off-box backups, and GPU senses behind one LiteLLM front door. The build, the honest numbers, and why the 'GPU node' runs on the CPU today.

DGX SparkTuring PiLocal AIBuild in Public
August 31, 20266 min read

Self-Hosted AI Image Generation: A 45 KB Hero

Self-hosted AI image generation on our DGX Spark made brew-space's launch hero — a reproducible, commercially-licensed 45 KB WebP at zero marginal cost.

Self-Hosted AIDGX SparkImage Generation
August 24, 20265 min read

The benchmark was green and the agents were broken

Our LLM benchmark said 0.84 while agent runs failed. The culprit: speculative decoding corrupting vLLM output — found with a $0 determinism probe.

vLLMBenchmarksCoding Agents
August 23, 20267 min read

The zero that wasn't a bug: DGX Spark code benchmark results

HumanEval, BFCL and MBPP on a DGX Spark: a 33B coding specialist vs a 35B generalist, the parallel tool-call zero we chased to the model, and the caveats.

DGX SparkBenchmarksCoding Agents
August 22, 20267 min read

DGX Spark code benchmarks, a BFCL bug caught early, and Ox Alpha

We added HumanEval, MBPP and BFCL to our DGX Spark benchmark lane, caught the BFCL bug that rewards never calling a tool, and tried free Ox Alpha.

DGX SparkBenchmarksCoding Agents
August 12, 20265 min read

Nemotron 3.5 Lightning on DGX Spark: day-one benchmarks

NVIDIA's Nemotron 3.5 Lightning benchmarked on DGX Spark 24 hours after release — plus the Ollama regression and runaway-reasoning failure the harness caught.

Local LLMDGX SparkBenchmarks
August 8, 20265 min read

DGX Spark image + video studio: one day, real numbers

We built a local AI image + video studio on a DGX Spark in a day. Real GB10 benchmarks, the license traps verification caught, and the quality ceiling.

DGX SparkLocal AIBuild in Public
August 5, 20267 min read

Before you upgrade your self-hosted AI agent, contain it

A big release landed for our self-hosted AI agent. We contained it first — and found two systemd directives that don't do what they look like.

AI GovernanceSecurityBuild in Public
July 29, 20265 min read

How much traffic hits an unlaunched website? We checked

Our unlaunched website served 14,000 requests to nobody. How to read pre-launch traffic, pick cookieless web analytics for $0, and the trap we found.

Web AnalyticsPrivacyBuild in Public
July 26, 20266 min read

DGX Spark round 2: two 120B MoEs and the number that lied

DGX Spark benchmarks for gpt-oss-120B and Laguna-S-2.1 — first independent numbers — plus the silent GPU bug that corrupted our published results.

Local LLMDGX SparkBenchmarks
July 24, 20266 min read

Our DGX Spark bake-off: 100 tok/s from a 35B MoE, and the dense-70B trap

Real DGX Spark local LLM benchmarks: a 35B MoE at ~100 tok/s, why dense 70Bs are a trap on unified memory — with corrected serving-stack numbers.

Local LLMDGX SparkBenchmarks
July 23, 20265 min read

We sent 29 AI agents to attack our own pitch deck

Adversarial AI review, applied to our own pitch deck: 29 agents attacked every claim. All six moat claims survived only after being rewritten honest.

AI AgentsAdversarial ReviewBuild in Public
July 12, 20265 min read

We built an AI chief of staff. It could run code.

We built an AI chief of staff meant to have one tool. AI agent security bit us: 'allowed tools' didn't restrict anything. Here's the fix.

AI AgentsSecurityBuild in Public
July 6, 20265 min read

Connecting Slack to Copilot Studio: what the docs don't say

We built a Slack-to-Copilot Studio relay in one night. Five things the docs don't tell you — auth, billing, and one SDK bug that matters.

Copilot StudioSlackAzure
July 5, 20266 min read

Our mascot was designed, judged, and vetted by AI agents

Multi-agent panels, an adversarial trademark sweep, a licensing-audited pipeline: our mascot, made in an afternoon for $0 in new software spend.

AI AgentsBrandingBuild in Public
July 4, 20263 min read

Why we run OpenTofu (and what its lock file taught our CI)

OpenTofu is a drop-in replacement for Terraform — except for one sharp edge in the lock file that our CI found before we did.

OpenTofuArchitectureTerraform
July 1, 20262 min read

From GoDaddy page builder to Azure: rebuilding rawrtech.ai

How we moved our own site from a template builder to versioned infrastructure on Azure with Cloudflare DNS — practicing what we preach.

AzureIaCMeta
June 18, 20262 min read

Shipping LLM features that survive production

The gap between a great demo and a dependable product is wider for AI than for any software we've built before. Here's how we close it.

LLMsEngineeringEvals
May 27, 20262 min read

RAG quality is a data problem, not a model problem

When retrieval-augmented generation disappoints, teams reach for a bigger model. Nine times out of ten, the fix is upstream — in the data.

RAGDataArchitecture