TEDWARD.NET · A WEEKLY DEV BLOG BY TED WARD
Dispatches from the sharp end of AI & software.
SHOWING 7 / 71 DISPATCHES — TOPIC: ALL
ALL 71 POSTS ↗
Decisions, Not Strings: Why TypeSafe's Jev Changes System Architecture
For years, developers have forced Large Language Models to act as glorified decision engines wrapped in fragile schema parsers. TypeSafe's Jev introduces System One models designed for structured, typed decisions rather than token generation. Here is why this architectural shift matters for software engineering.
The Generational Fumble: Why Google Is Retreating Behind the Walled Garden
Google held every advantage in the AI revolution: custom TPUs, the original transformer architecture, and bottomless ad profits. Here is why they fumbled developer mindshare and are retreating behind enterprise lock-in.
The LLM Plateau: When the Model Stops Being the Moat
As the brute-force scaling of LLMs encounters economic realities, the primary AI moat is shifting from raw model intelligence to context and coordination layers. Discover why context is the new source code and how to build decoupled, resilient AI architectures.
From Typist to Architect: How to Teach Software Engineering in the Age of AI
In the era of generative AI, teaching syntax line-by-line is obsolete. Discover how to mentor junior engineers on architectural patterns, separation of concerns, and guiding AI to produce robust, scalable code.
The Regulatory Moat: When the Frontier Becomes a Cartel
The recent viral resignation of an Anthropic researcher revealed more than safety concerns: it exposed an attempt by Silicon Valley insiders to build a regulatory moat around AI. Learn why sovereign, local-first architecture is our only defence against an emerging AI cartel.
Why Agentic Coding Won't Make Software Developers Extinct
A recent LinkedIn post claimed agentic AI coding will turn software development into a lost art like blacksmithing. Here is why that argument completely misses how technical abstractions actually work.
Being My Own First Customer: What Dogfooding the Rebuilt Foundry Actually Broke
When pointing the newly rebuilt Foundry at a real-world, 5,574-file monorepo, the context graph did not gracefully scale - it broke. Here are the 25 bugs, quadratic bottlenecks, and silent hallucinations uncovered by dogfooding an AI context substrate.