FROM THE TRENCHES Issue №310

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.

GDPR plus EU AI Act stacking for DACH SaaS founders Business & Regulation

GDPR + EU AI Act for DACH SaaS

A current framework for classifying AI use cases, mapping GDPR and AI Act duties, assigning evidence owners, and planning applicable deadlines.

Evidence-based review of why AI agent projects are paused or cancelled AI & Agents

Why AI Agent Projects Get Cancelled

What Gartner’s forecast does and does not show, plus an evidence-based review of value, evaluation, tools, cost, handoff, data, scope and governance risks.

RAG versus fine-tuning versus long-context architecture decision in 2026 AI & Agents

RAG vs Fine-Tuning vs Long-Context 2026

A workload-based method for testing retrieval, supervised fine-tuning, long context, and hybrid architectures without relying on universal cost crossovers.

LLM API cost modeling and architecture decisions in 2026 AI & Agents

LLM API Costs 2026. Architecture Shift

A current method for modeling input, output, reasoning, caching, tools, batch work, retrieval, and operations before changing an AI architecture.

Ethereum to Solana migration cost teardown Web3 & Privacy

Ethereum to Solana Migration Cost

Scope an Ethereum-to-Solana migration from program semantics, accounts, assets, wallets, data, operations, security, state transition, and user continuity.

Zero-knowledge proof use cases outside crypto Web3 & Privacy

Zero-Knowledge Outside Crypto

Six non-crypto ZK use cases. Privacy-preserving KYC, age verification, supply chain provenance, private credentials, confidential ML inference.

Account abstraction ERC-4337 in production Web3 & Privacy

Account Abstraction in Production

What ERC-4337 enables, what depends on the wallet implementation, how EIP-7702 changes EOAs, and which bundler, paymaster, recovery, and upgrade risks to test.

Road to PMF Product & MVP

A Measurable Road to Product-Market Fit

Product-market fit is not a badge or a single survey score. Define a market and use segment-specific evidence from activation, retention, willingness to pay, acquisition, and service economics.

Feature fabric Product & MVP

Escaping the Feature Trap with Evidence

Feature volume is not a product strategy. Learn how to connect backlog decisions to user evidence, product goals, risk, operating cost, and measurable outcomes.

Inbox, without the noise

Get the next field note

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