In February 2026, as AI agents like OpenClaw go viral, the productivity puzzle that has vexed economists for decades feels closer to cracking. Yet most users still treat AI as a glorified search engine: querying facts from a probability machine that hallucinates as often as it enlightens. This misses the point entirely. The real leap comes from tinkering: building, iterating, and deploying AI augmented products, services, and personal agents that amplify human work. Do so, and you future proof your role in an economy where jobs evolve or evaporate. Ignore it, and you risk becoming irrelevant in a world where AI turns ordinary workers into supercharged creators or leaves them behind entirely.
The story of OpenClaw illustrates the shift. It’s an AI agent created by Austrian developer Peter Steinberger who is a solo programmer and not associated with Big Tech. This open source framework exploded to two million visitors in its first week, claiming the fastest GitHub growth on record. What sets it apart? AI agents with persistent memory, contextual awareness, and infused personalities that make them feel almost alive.
Steinberger built much of it using the agent itself, a self bootstrapping feat that turned heads. Users aren’t just chatting; they’re prompting agents to manage schedules, write daily journals, or even role play as quirky companions. This framework integrates existing big tech LLMs with your computer’s day to day tasks which is why most people just install it on a Mac Mini to tap into their existing calendars, communication and other, mostly personal data. It’s not sentient AI yet but it’s a glimpse of tools that evolve with you, democratising creation in ways corporate labs have struggled to match.
Economists have long pondered why technological marvels deliver underwhelming gains. The Industrial Revolution took generations to boost living standards; computers and smartphones added modest productivity bumps, often delayed by adoption lags. AI could be different. Unlike static tools, agentic systems like OpenClaw learn from interactions, handling repetitive tasks with nuance. OECD data shows AI exposed sectors could see productivity gains of 0.25 to 0.6 percentage points annually over the next decade (gains this author finds highly understated), with wages potentially rising in complementary roles. The catch? Most people misuse it by taking outputs at face value rather than probing, refining, or integrating. It’s a likelihood engine, not an oracle; even the newest models remain probabilistic at their core, predicting next tokens based on patterns rather than true understanding, though their coherence has improved dramatically. We argue that the magic only emerges when you tinker.
Consider the evidence. Programmers, once the artisans of the digital age, now risk becoming the conveyor belt workers of the 21st century. Anthropic claims 80% of its latest model, Claude Opus 4.6, was coded by itself. Steinberger’s OpenClaw bootstrapped its own development. Even novices can ship apps in hours: one tinkerer (this writer included) deployed a simple tool from concept to production without prior expertise, guided by Claude Code. These aren’t anomalies; they’re harbingers. McKinsey forecasts 85 million jobs displaced by 2030, but 97 million new ones created, if we manage to adapt.
At first glance, the fear is that AI steals the fun stuff such as creativity and ideation leaving drudgery like household chores. Yet as David Mattin argues in his prescient thread What Do I Tell My Children (link), the transformation runs deeper. AI excels at codable, scalable tasks; humans thrive in empathy, narrative, and the unpredictable. The economy shifts toward micro entrepreneurship, where individuals orchestrate AI fleets for personalized value. The act of creativity itself may evolve, as Mattin explores in his thread turned essay. AI can generate ideas at scale, but true innovation often sparks from human friction—the messy interplay of intuition, collaboration, and serendipity that no algorithm can fully capture. Tinkering with AI doesn’t diminish this, but it amplifies it by turning creators into conductors who blend machine efficiency with human insight.
Peering into 2035, the outlook is brighter than the dystopias suggest. Imagine a world where AI agents handle logistics, data crunching, and routine decisions, and thereby freeing us humans for what machines can’t replicate: the subtle alchemy of face to face connection. Psychotherapy? An agent might diagnose patterns flawlessly, but the pheromone magic, in which subconscious trust is built through scent, micro expressions, and shared space, remains uniquely human. In person sessions become premium experiences, spawning new models like empathy retreats or bio enhanced social hubs where insightful humans command premiums.
In such a future reality, human to human interactions will appreciate in value, proliferating business models around authenticity. Think artisanal conversations: networks of storytellers curating unscripted events, presence coaches blending psychology with VR hybrids, or community curators fostering bonds in a hyper digital age. Yet this pivot extends to the hard edges of industry where AI streamlines production of industrial goods, optimises transportation fleets with predictive logistics, bolsters defence through autonomous simulations, and accelerates chemistry breakthroughs via molecular modelling. Jobs evolve into orchestrators overseeing AI ethics, designing narratives for agent personalities, or auditing human AI harmony in factories, labs, and supply chains. Wealth concentrates less in code and more in connection; micro communities flourish, turning social capital into economic engines. It’s not utopia but a positive pivot where technology elevates the human element.
There is a catch though, for example, if inequality lingers and skills gaps widen. On the other hand, large corporations and institutions will lag, bogged down by legacy systems, regulatory hurdles and data privacy concerns, and the sheer societal stakes: ethical biases, job churn and privacy erosions. Adoption will not be uniform; economic dead zones could emerge in hubs slow to tinker. Normal people in admin jobs, clerks, finance, auditors, and accountants face particular risks. They will not be able to pivot overnight, but with accessible tinkering tools, even they can upskill by automating rote tasks and focusing on oversight or client relations.
Yet the way forward is clear: Start experimenting. Download Claude Code, Codex or even OpenClaw: build a simple agent to journal your day or automate a workflow. Tinker not to replace your job, but to reinvent it. In an unforgiving decade of disruption, the future belongs to those who treat AI as a collaborator, not a crutch. The style may be experimental; the substance is empowerment.
Thanks for reading The Friday Thinkers! Subscribe for free to receive new posts and support our work.
Yours,
M.
The Friday Thinkers