Author

Linyb Geek Road

Tech notes

181

Articles

Likes

1.1k

Views

Comments

Latest from Linyb Geek Road

100 recent articles max

Linyb Geek Road

Sep 29, 2026 · Artificial Intelligence

Reverse-Engineering Jev: 10K API Calls Expose Closed-Source Model Architecture

An independent researcher reverse-engineered TypeSafe's closed-source Jev classification model using 10,000 API calls, revealing its architecture uses a classification head with shared prefix inference, listwise scoring, and exceptional calibration (ECE 0.0313), challenging assumptions about black-box security.

JevMixture of Expertsblack-box probing

0 likes · 22 min read

Reverse-Engineering Jev: 10K API Calls Expose Closed-Source Model Architecture

Linyb Geek Road

Sep 28, 2026 · Artificial Intelligence

Jev Decision Model: Extracting Structured Judgments from LLMs for Code-Driven Routing

This article details the Jev decision model from TypeSafe, which provides typed answers (Choice, Score, Noul) with probabilities instead of text, enabling code-driven routing, verification, and retrieval reranking with lower latency and cost than LLMs, while explaining its four architectural patterns, confidence gating, atomic decomposition, and limitations.

AI-powered softwareJevLLM verification

0 likes · 25 min read

Jev Decision Model: Extracting Structured Judgments from LLMs for Code-Driven Routing

Linyb Geek Road

Sep 26, 2026 · Artificial Intelligence

Graphify: One-Command Knowledge Graphs for AI Coding Assistants (120k Stars)

Graphify builds queryable knowledge graphs from codebases using local tree-sitter parsing for code and LLMs for docs, enabling AI assistants to traverse real code relationships via query, path, and explain commands, with benchmark results showing 0.497 recall@10 on LOCOMO versus 0.048 for mem0.

AI coding assistantGraphifyLOCOMO benchmark

0 likes · 9 min read

Graphify: One-Command Knowledge Graphs for AI Coding Assistants (120k Stars)

Linyb Geek Road

Sep 24, 2026 · Artificial Intelligence

How Jev Cuts Agent Orchestration Costs 80% with Dynamic Harness Generation

Jev, a specialized classification model, reduces agent orchestration costs by 80% by handling routing, evaluation, intent decomposition, and dynamic harness generation at a fraction of GPT-4o's cost, enabling self-orchestrating multi-agent systems while maintaining quality, though plan-level error detection remains an open challenge.

Cost OptimizationJevLLM routing

0 likes · 12 min read

How Jev Cuts Agent Orchestration Costs 80% with Dynamic Harness Generation

Linyb Geek Road

Sep 23, 2026 · Artificial Intelligence

Jev: TypeSafe's Decision Model for High-Frequency Agent Tasks

This article analyzes TypeSafe's Jev, a specialized decision model that returns structured choices, scores, or boolean judgments for high-frequency agent tasks like game AI, browser automation, and ticket triage, offering lower latency and cost than LLMs by narrowing training to calibrated classification with confidence gating.

JevLLM alternativeTypeSafe AI

0 likes · 17 min read

Jev: TypeSafe's Decision Model for High-Frequency Agent Tasks

Linyb Geek Road

Sep 20, 2026 · Artificial Intelligence

Two AI Diagram Skills That Supercharge DeepSeek for Architecture Visuals

The author demonstrates how two open-source skills, fireworks-tech-graph and architecture-diagram-generator, enable DeepSeek to generate professional architecture diagrams in multiple styles from natural language prompts, eliminating manual drawing effort.

AI-assisted designArchitecture DiagramsDeepSeek

0 likes · 7 min read

Two AI Diagram Skills That Supercharge DeepSeek for Architecture Visuals

Linyb Geek Road

Sep 19, 2026 · Backend Development

Archify: AI Agent Skill Generates Validated Interactive Architecture Diagrams from Code

Archify is an AI agent skill that analyzes code repositories to produce validated, interactive architecture diagrams in five types — Architecture, Workflow, Sequence, Data Flow, and Lifecycle — with a rigorous JSON Schema and layout validation pipeline, an interactive HTML viewer supporting search, path tracing, role comparison, and Architecture Delta for PR diffing, all delivered as a single shareable HTML file.

AI AgentArchifyArchitecture Diagrams

0 likes · 11 min read

Archify: AI Agent Skill Generates Validated Interactive Architecture Diagrams from Code

Linyb Geek Road

Sep 17, 2026 · R&D Management

Practicing AI Specs with OpenSpec and OpenCode: A Step-by-Step Tutorial

This tutorial demonstrates how to use OpenSpec and OpenCode for spec-driven AI-assisted development by building a fund real-time valuation desktop app with Tauri, covering project planning, UI design with Pencil, OpenSpec's three-phase workflow, and incremental feature implementation.

AI SpecsOpenCodeOpenSource Tools

0 likes · 18 min read

Practicing AI Specs with OpenSpec and OpenCode: A Step-by-Step Tutorial

Linyb Geek Road

Sep 16, 2026 · Artificial Intelligence

Merging OpenSpec & Superpowers: 7 Integration Pitfalls & the Bridge Skill That Fixes Them

The author details seven integration pitfalls when combining OpenSpec's specification-driven development with Superpowers' disciplined TDD workflow, demonstrates a real Go microservice JWT authentication case, and provides a custom bridge Skill to automate spec-to-plan translation, dual review, and verification.

AI-assisted developmentGoJWT authentication

0 likes · 29 min read

Merging OpenSpec & Superpowers: 7 Integration Pitfalls & the Bridge Skill That Fixes Them

Linyb Geek Road

Sep 13, 2026 · R&D Management

Code Rewrites: When Programmer Ego Trumps Business Value — 4 Decision Criteria

The article argues most rewrites satisfy programmer aesthetics, not business needs; stable legacy code encodes hard-won bug fixes; rewrites risk reintroducing old bugs; four concrete criteria justify rewrites: unsupported dependencies with vulnerabilities, knowledge silos, measurable productivity decline, or business outgrowing architecture; AI lowers rewrite cost but not the risk of lost context.

AI-assisted developmentJoel Spolskybusiness value

0 likes · 9 min read

Code Rewrites: When Programmer Ego Trumps Business Value — 4 Decision Criteria