McLuhan Digital Tetrad

What it does

• Runs McLuhan's four laws against any technology you name, tuned by target year, time scope, impact scale, depth, and how speculative to be.
• Renders the result as a figure-and-ground grid of flip cards: the analysis on the front, the counterargument on the back.
• Expands any quadrant into a deep dive, and any open question into a long-form, cited answer.
• Adds a tools layer: a McLuhan persona to debate, a product-ideas generator, and a disruptive-companies card.
• Exports the whole reading to a consolidated PDF report.
• Runs on Claude with a local Ollama failover, so it keeps working when the network does not.

Context

Marshall McLuhan argued that every technology does four things at once. It enhances something, it makes something obsolete, it retrieves something older, and pushed far enough it reverses into its opposite. Most people meet the tetrad as a diagram in a theory class. This app treats it as a tool. Type in any technology, set the lens you want, and it runs the full four-law analysis, then keeps going: it argues with its own findings, opens any thread into a long-form answer, lets you debate a simulated McLuhan, and turns the reading into product ideas and a list of the companies already moving on it. The same tetrad engine runs as phase two of the What If? Magazine pipeline. A standalone instrument here, operationalized there.

What was built

A Next.js and TypeScript app with a real control surface, not a single prompt box. An Analysis Parameters panel tunes the read: a technology to analyze, a target year, and sliders for time scope (immediate to long-term), impact scale (individual to societal), depth (practical to philosophical), and how speculative to be. Each setting is translated into natural-language guidance that shapes the prompt rather than a raw knob.

The core output is the tetrad itself: three concrete effect points per law, essay-style considerations, a synthesis summary, a confidence score, and the classic McLuhan grid rendered as figure-and-ground. Every quadrant is a flip card. The front carries the analysis, the back carries the counterargument, so the tool disagrees with itself by design. Deep Dive expands any quadrant into a worked example and open questions, and any question expands again into a long-form, cited answer on demand.

Around the analysis sits a tools layer: a context-aware McLuhan persona you can debate, a product-ideas generator that derives concepts from the reading, and a disruptive-companies card that surfaces real players with funding and founding dates. A consolidated report walks from tetrad to deep dive to summary and exports to PDF.

Underneath, the resilience story is the point. It runs Claude (claude-opus-4-8) as the primary provider through the Anthropic SDK, with a local Ollama model (Qwen2.5-coder 32B) as automatic failover or an explicit toggle, both held to schema-constrained structured outputs so the JSON never drifts. Six API routes serve analysis, exploration, the chatbot, product ideas, and per-question deep dives, each with graceful fallback content on error. The Ollama path runs on the same local-first stack documented in Local Frontier.

What was learned

The hard lesson came before the features. An early build depended on a single upstream provider and stalled whenever that dependency went down, so the app limped along on hardcoded simulated responses. The fix was architectural, not persistence: treat the primary cloud model as one option, not the only one. The rebuild on Claude with a local Ollama fallback and schema-constrained output on both paths did not throw the workaround away. The simulated-response path matured into the failover architecture, so the thing that was once a dead end became the safety net.