Categories
AI
13 posts

I Tried to Predict Bitcoin From Sunspots, Then Built a Startup on the Leftovers
How I killed a bad trading idea with proper statistics, then skipped the same rigor for the GPS business I built out of its data.
aiarchitecture
My First RAFT Fine-Tune on a Shared 16 GB GPU: 7 Gotchas the Paper Doesn't Mention
I fine-tuned Qwen3.5 with RAFT to answer in one creator's voice with verified citations. The training was the easy part.
aibenchmarksllmStop Preparing for AI Job Loss Start Preparing for AI Meaning Loss The Real Crisis Isn't Unemployment—It's What Happens When Work No Longer Defines Us
Why the AI meaning crisis poses a deeper threat than technological unemployment, and how to prepare for a future where work no longer defines human purpose.
aiphilosophy
Your Code Knowledge Graph Needs Four Layers, Not One
Memory should be a graph. The graph should be four layers. The middle one carries navigation. The top one turns navigation into action.
agentsaiarchitecture
Streaming UI from AI Agents: 5 Tools Ranked
AG-UI, A2UI, Vercel AI SDK, TanStack AI, and Kombai take different bets on how agents should build interfaces. The architectural question decides which wins.
aiagentsfrontendarchitecture
Capital-of Is Not a Single SAE Feature. So I Built a Mutation Loop to Find What Is.
SAE features can't isolate relations in Gemma-2-2B. I built a mutation-selection loop that can. The bottleneck was tokenization.
aiinterpretabilityllmsparse-autoencoders
Your MCP Setup Is Burning 90% of Its Context Window. Here's the Fix.
Every MCP tool you connect loads its full schema upfront, before you type a word. Anthropic's deferred loading fixes this.
mcpclaudeaiagents
Sparse Autoencoders Can't Measure Generation-Time Behavior. That's Not a Bug.
Why sycophancy SAE features have Cohen's d=9.9 but hallucination detection fails. The answer turned out to be deeper than measurement timing.
aiinterpretabilitysparse-autoencoders
Svapna or Sushupti: What Three Traditions Say About Offline Memory Consolidation
Neuroscience, recent AI papers, and an ancient Sanskrit text converge on the same insight about offline consolidation, yet disagree on which phase of sleep matters most.
aiagentsmemorymuninndb
Stop Designing Your AI System. Design Its Environment.
Self-evolving AI harnesses fail when they optimize a fixed evaluator. The biological model is right: what needs to evolve is the selection pressure, not just the genome.
aiagentsarchitectureinterpretability
Your Agent's Memory Metrics Are Lying to You. Here's How to Ground Them.
Memory consolidation looks great on dashboards. But if your metrics can improve without retrieval getting better, you're optimizing a detached proxy.
aimemorybenchmarksmuninndb
Why AI Agents Need Sleep
AI agents capture memories but never consolidate them. Here's how MuninnDB's Dream Engine borrows from neuroscience to fix that.
aimemorymuninndbarchitecture
Stop Putting AI in Your Apps. Put Your Apps in AI.
Why AI features inside traditional apps are backwards, and how MCP flips the architecture.
mcpaiarchitecture