The solution might be cancelling my AI subscription

AI, learning, and craftsmanship

• Strong split on whether AI-assisted coding yields real learning.
• Critics say delegating implementation to models is like blindly following GPS: you reach the destination but don’t build the underlying map, so skills atrophy.
• Others argue there is deep learning in using AI well: long-horizon workflows, specs, tests, multi-agent systems, model comparisons, security of agents, etc.
• Some think most of that is fleeting, model-specific trivia that will be obsolete fast; better to invest in timeless engineering skills.

ADHD, attention, and dopamine

• Many with ADHD describe AI as an “amplifier”: it makes it trivial to spin up endless projects, fragment attention, and chase dopamine instead of shipping or maintaining anything.
• Others with ADHD report the opposite: AI provides structure, externalized executive function, and lets them finish or stay focused on a single main product.
• Broad agreement that AI chat/agents can create “pseudo-productivity”: lots of busy interaction that feels productive but may be slower than reading good docs or doing deep work.

Side projects, value, and meaning

• Debate over whether a long list of AI-generated apps is joyful exploration or a depressing pile of shallow, disposable slop.
• Some see hobbyist tinkering as inherently fine, like LEGO or crosswords; not everything needs a business model.
• Others worry that ultra-cheap creation erodes commitment; without friction, there’s less pride, depth, and long-term significance.
• Several emphasize focusing on one substantial, meaningful project (or “needle”) rather than many unaligned experiments.

Productivity, quality, and maintenance

• Supporters say AI is superb for boilerplate, small one-off tools, infra scripts, and rapid prototyping; they cite real-world wins (e.g., migration utilities, home infra, personal dashboards).
• Skeptics note maintenance headaches: LLM output is unreliable, hard to review, often sloppier than hand-written code; monitoring it is exhausting compared to trusting a compiler.
• Concern that AI makes it easy to create large, untested codebases that are hard to understand without ongoing model access.

Tools, friction, and media lens

• Several frame AI like calculators or GPS: it removes “incidental friction,” freeing time for true ambiguity—but also removes learning opportunities and serendipitous skill-building.
• The distinction between incidental vs. meaningful friction is seen as user- and goal-dependent.
• Some use a McLuhan-esque view: AI doesn’t just extend cognition, it can replace parts of it, raising questions about attention, wisdom, and long-term craft.

Ethics, economics, and incentives

• Some call generative AI “stealing other people’s work,” energy-intensive, and wealth-concentrating; others ignore or downplay these issues.
• A recurring theme: tools are force multipliers, but “there has to be some force to multiply” and incentives still dominate outcomes. AI alone won’t create a viable business or meaningful work.