Scaling LLM Classification via Hallucination and Vector Search
Use LLMs to generate hypothetical categories and resolve them to a real taxonomy via vector similarity for cheaper, more scalable classification.
Structured Output
Techniques and tools for making LLMs reliably generate structured data formats like JSON, using grammars, schemas, token masking, and constrained decoding.
Reading List
Stop Humanising LLM Instructions; Start Rendering Them
Humanising LLM outputs at the instruction level sacrifices technical precision for prose and should be handled at the UI boundary instead.
The Slop Harness: Why SOTA Models are Regressing in Tool Adherence
228
Advanced LLMs are becoming less reliable at following general tool schemas because they are being over-optimized for specific, forgiving internal harnesses.
Interfaze: A Hybrid Architecture for High-Accuracy Deterministic AI
164
Interfaze is a hybrid AI model that merges DNN precision with transformer flexibility to outperform generalist LLMs in high-accuracy, deterministic tasks.
The Case for Local-First AI
1903
Developers should prioritize on-device AI over cloud APIs to create more private, reliable, and cost-effective software.
Guide to Chrome's On-Device Prompt API and Gemini Nano
277
The Prompt API enables web developers to integrate Gemini Nano for on-device, multimodal AI processing directly within the Chrome browser.
Lightfeed Extractor: LLM-Powered Web Scraping Library
A TypeScript library for robust, LLM-powered web data extraction and browser automation.
From Sampling to Grammars: Making LLMs Reliably Output Structured Data (Even for Thinking Models)
234
Use efficient sampling plus grammar constraints to guarantee format today, but expect models to natively emit structured outputs tomorrow—especially when you let them think first, then constrain.