Observed Signal · Oct 8, 2026 · Policy Update · Source: UX Collective · Impact: 2/5 · Sentiment: Neutral
Tesler's Law in AI: Complexity Moves, Not Disappears
This article examines how generative AI has shifted rather than eliminated complexity in software design, applying Larry Tesler's Law of Conservation of Complexity. It argues that AI interfaces like chat boxes transfer the burden of specification and verification to users and teams, often hidden by the apparent simplicity. Examples include a METR study showing a 19% productivity slowdown for experienced developers using AI, and Stanford RegLab findings on hallucination rates in legal AI tools. The piece highlights where AI genuinely absorbs complexity (e.g., customer support) and introduces a 'deterministic floor' for tasks requiring exactness. It concludes with actionable guidance for designers to make complexity allocation explicit and measure the true costs of AI adoption.
The article provides a conceptual analysis of AI's impact on UX and complexity, relevant to AdTech/MarTech for understanding AI adoption challenges and design implications, but it is not breaking news and focuses on theory rather than specific industry events.
Track Apple Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Larry Tesler's Law of Conservation of Complexity states that irreducible complexity can only be transferred, not eliminated.
- METR's July 2025 RCT with 16 developers found a 19% slowdown when using AI, contrary to expected speedup.
- A Stanford RegLab study found LexisNexis and Thomson Reuters legal AI tools hallucinated on 17-33% of queries.
- Stanford HAI's 2026 AI Index reports 70% of organizations use generative AI in at least one business function.
- A 2025 study of 5,172 customer support agents showed AI improved issues resolved per hour by 15%, especially for novices.
Connected Companies & Entities
4 Entities mapped“Steve Jobs recruited Tesler to Apple in 1980, where he rose to chief scientist....”
“METR ran a randomized controlled trial with developers to measure AI's impact....”
“Mentioned as a platform developer example from the past....”
“Thomson Reuters also built legal AI tools tested by Stanford RegLab....”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Avoid Overpaying for Unnecessary AI Complexity
The article explains that enterprises often apply overly complex AI architectures to simple marketing tasks, driving up total cost of ownership and verification overhead. It defines four mechanisms—rule-based, predictive, generative, and agentic—ranked by complexity, cost, and risk. The piece highlights examples and vendor/pilot pitfalls, cites EY analysis that agentic workflows raised per-interaction costs from about $0.04 in 2023 to roughly $1.20 in 2026, and references Gartner estimates that agentic tasks can use 5–30× more tokens than standard genAI chatbot interactions. The author recommends choosing the lightest mechanism that meets requirements, pricing pilots at production volume, and asking vendors for mechanism-level cost estimates including review and verification costs.
AI Agents Increase Work — Verification Becomes Key
At Fortune Brainstorm Tech executives from multiple companies warned that AI agents are producing significant amounts of work but creating new verification and accountability burdens. Examples cited include an Openclaw agent that deleted a researcher’s emails and reports that generated code often requires heavy revision. Speakers — including leaders from May Mobility, Trustguard AI, Thomson Reuters and Sentinel One — argued for greater transparency, separated verification systems, and self‑regulating or cross‑checking agent architectures to reduce risky errors. Survey data referenced shows many employees see no time savings from AI, while some leaders report material time gains; the industry is searching for automated, safety‑centric validation methods used in critical systems to scale verification efforts.
AI Agents May Slow Development and Harm Quality
The article argues that while AI agents and coding tools can increase engineering output, they may simultaneously reduce product quality, introduce outages, and create long-term technical debt. It cites examples: Anthropic’s Claude-powered development (reportedly 80%+ of production code) shipped a persistent UX bug that affected paying users until public complaint prompted a fix; Amazon experienced outages tied to AI-assisted changes (AWS reported a 13-hour interruption after an agentic tool deleted and recreated an environment), triggering mandates for senior sign-off on junior AI-assisted changes; and large firms (Uber, Meta) are using AI-usage metrics in performance assessments, pressuring engineers to adopt agents. Startups and researchers report short-lived velocity gains followed by maintenance burdens. The piece recommends stronger architecture, formal validation, and renewed QA practices to manage agentic risks.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.