FROM THE TRENCHES Issue №354

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.

OpenSandbox review illustration with a shield representing the boundary around an AI agent workload AI & Agents

Alibaba OpenSandbox: Self-Hosting and Production Limits

A version-aware OpenSandbox review: Docker, Kubernetes, credentials, tenant pools, artifact verification and a practical production pilot.

Transformers.js running private local AI in a browser with WebGPU and WebAssembly fallback AI & Agents

Transformers.js Browser AI: When Local Inference Belongs in Your Product

A production guide to private in-browser AI: verified adoption, WebGPU and WASM, real cost, privacy boundaries, use cases, fallbacks and a 30-day pilot.

Cloudflare Kitesurf isolate beside a Chromium browser fallback lane AI & Agents

Cloudflare Kitesurf Review: Cost, Limits and Production Fit

Kitesurf uses less CPU and memory than Chromium but trades speed and compatibility. Review beta pricing, security limits, alternatives and a 30-day pilot.

Opportunity map for AI, software and automation in grocery retail Delivery & QA

AI in Grocery Retail: An MPREIS Opportunity Map

Independent public-source MPREIS analysis: hypothetical grocery-retail AI proposals and an illustrative pilot sequence. No company deployment or ROI is claimed.

A stateful LLM platform connected to persistence, access control, background jobs, APIs, deployment and QA gates Delivery & QA

Stateful LLM Platform Production Architecture

An anonymized field guide to persistence, tenant access, long-running jobs, external interfaces, deployment controls and continuous QA without a full rewrite.

Semantica decision provenance graph connecting evidence, policy gates and an AI agent outcome Business & Regulation

Semantica Review 2026: Can It Explain Every AI Agent Decision?

A buyer-focused review of Semantica for decision provenance, policy gates and regulated AI audit trails, including security limits and a two-week pilot scorecard.

A layered healthcare operations map connecting patient services, clinical workflows, automation and AI governance Delivery & QA

Tirol Kliniken (TILAK): A Digital Operations Opportunity Map

Independent public-source Tirol Kliniken analysis: hypothetical software and AI proposals. The pilot sequence is illustrative, not clinical validation or a delivery promise.

Layered digital energy services connecting customer journeys, deterministic controls, AI assistance and human operations Delivery & QA

TIWAG Digital Energy: A Software and AI Opportunity Map

Independent public-source TIWAG analysis: hypothetical digital-energy proposals. No commission, internal-system assessment or measured company outcome is claimed.

A structured specification turns a vague AI coding prompt into reviewed requirements, a plan and testable tasks AI & Agents

GitHub Spec Kit Review: Is It Worth the Process?

GitHub Spec Kit replaces one-shot vibe coding with reviewed specifications, plans and tasks. Here is where that structure pays off, where it adds ceremony and how to run a production pilot.

Inbox, without the noise

Get the next field note

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

Facing one of these decisions right now? Talk it through with us →