AI-native systems
How AI fits into real workflows without turning everything into a black box — especially systems where AI assists, but deterministic logic still owns the final action.
- A Web Interface for the NHTSA Evidence System
An NHTSA evidence interface limits users to eight validated questions, separates record types, and labels approximate retrieval with similarity scores.
- Every Answer Cites Its Source
A Microsoft Fabric evidence system routes five NHTSA question types through deterministic lookup and three through cited semantic retrieval.
- What a Six-Language Horoscope Pipeline Taught Me About Localization
I automated horoscope generation, speech, video assembly, and publishing in six languages. Translation scaled; register, voice, and cultural expectations did not.
- What AI Crawlers Could Read on My Site
Scanning the same pages for SEO and answer-engine visibility exposed meaningful differences in crawler access, parsing, and citation readiness.
- What Changed When Models Could Answer Each Other
I compared parallel model responses with a sequential, stateful debate in which each model could rebut or concede points made in earlier turns.
- I Built Automatic Backlinks Before Testing the Writing
A writing tool extracted entities and contradictions, but months of data-model refinement happened before testing whether it improved a real manuscript.
- Building Durable Memory Around a Stateless Model
A durable memory design stores facts and narrative history outside the model, retrieves them differently, and reviews every proposed update.
- You Can Check a Cookie
An AI cookie recipe never set, while established recipes did. Ingredient ratios provided the deterministic check that plausible prose could not.
- Three Graphics Bugs in an AI-Written WebGL App
An AI-generated three.js app ran immediately but mishandled glyph winding, bevel normals, and post-processing alpha until visual checks exposed each bug.
- Why Evidence Weight and Agreement Need Separate Scores
A proof of concept that scores support independently from source disagreement, discounts duplicated evidence, and preserves unresolved conflicts.
- Four Results That Looked Better Than They Were
A GPU benchmark, an MRI classifier, an imbalanced split, and a text model all produced numbers that needed more scrutiny than celebration.
- Using a Netcode Pattern to Contain AI Systems
Client prediction, server authority, reconciliation, and event logs offer a practical analogy for placing probabilistic AI inside deterministic software.
- A Text-to-SQL Prototype for Patient Data
A patient-data text-to-SQL prototype needed semantic metadata to draft useful queries and a human review gate before anything could be trusted.
- What a 200-Line Keras Tutorial Left Out
A 3D-CNN tutorial ran in minutes; applying it to real brain MRIs took months and 25 GPU-hours to reach 61% accuracy against a 52% baseline.
- AI Is Good at the Work You Can Check
Coding assistants saved time when tests, schemas, or reconciliations could check their work; without those checks, plausible mistakes consumed the gain.