FROM THE TRENCHES Issue №308

Practical decisions for AI and software

Technical guides to choose models, verify software and plan delivery.

Latest articles

New comparisons, practical guides and engineering lessons.

A decision splits an incoming business request into a controlled workflow and a review queue AI & Agents

Laya and Jev for Business: 6 Practical Workflows

Six practical ways to reduce manual handoffs with Laya and Jev. Includes n8n integration, review-first code and an honest support-routing business case.

Kevin Riedl
Read
Wavect Pake editorial illustration showing layered web windows becoming a desktop app Delivery & QA

Pake: Turn Websites into Desktop Apps Without Electron

One command, a lightweight desktop shell. What Pake really saves, why a Linux AppImage can be 76 MB, and which login and deployment checks matter.

Kevin Riedl
Read
Wavect editorial graphic separating model performance from product defensibility in the Laya vs Jev comparison AI & Agents

Laya vs Jev: What the Benchmarks Mean for AI Startups

Can an open specialist undermine a two-year head start? What Laya actually shows, what the Jev comparison misses, and where a defensible AI business can create value.

Kevin Riedl
Read
Wavect editorial header for Google AX, with a shield representing agent execution boundaries AI & Agents

Google AX: Suspend/Resume, Restart Costs and Self-Hosting

Fast resume is not crash recovery. What Google AX preserves, what changed in its API, and how to test whether fewer replays justify the infrastructure.

Kevin Riedl
Read
Raw agent traces becoming a persistent wiki and a validated SKILL.md file AI & Agents

WikiSkill: How Agents Evolve SKILL.md from Experience

WikiSkill compiles agent traces into a persistent evidence wiki, then validates atomic changes to SKILL.md. We review the benchmarks, transfer results, limits and production architecture.

Kevin Riedl
Read
Octop connecting users, AI agents, tools and self-hosted data in one application stack AI & Agents

Tencent Octop Review: What Builders Actually Get for Free

Tencent released Octop under MIT, with multi-user agents, a dashboard, automation and storage. We examine what is really included, what is not and what a production fork still costs.

Kevin Riedl
Read
Wigolo fans one AI agent query across search engines, reranks evidence locally and exposes ten web tools through MCP AI & Agents

Wigolo Review: Local Web Intelligence for AI Agents

Wigolo gives AI agents local-first search, fetch, crawl and research without a paid search API. We verify the ten-tool surface, cache, privacy, AGPL license and production limits.

Kevin Riedl
Read
Wavect editorial illustration of a typed decision split for the Jev AI model AI & Agents

Jev AI Review: Decision Models for Agent Workflows

Jev makes typed decisions instead of prose. See how it can route models, govern agent control steps and support bounded automation without replacing hard rules.

Kevin Riedl
Read
Layered context cards representing persistent AI agent memory in Wavect editorial artwork AI & Agents

Supermemory: AI Agent Memory, RAG and Local Setup

Give your AI agent context between sessions without building the memory pipeline yourself. A practical Supermemory guide with API code, local setup and an honest benchmark review.

Kevin Riedl
Read
Inbox, without the noise

Get the next field note

One concise email when we publish. No tracking pixels, and no inbox filler.