AI Tools: Capability, Caution, and Critical Thinking
One morning, when Gregor Samsa woke from troubled dreams, he found himself transformed in his bed into a horrible vermin. He lay on his armour-like back, and if he lifted his head a little he could see his brown belly, slightly domed and divided by arches into stiff sections.
The bedding was hardly able to cover it and seemed ready to slide off any moment. His many legs, pitifully thin compared with the size of the rest of him, waved about helplessly as he looked. “What’s happened to me? ” he thought. It wasn’t a dream.
His room, a proper human room although a little too small, lay peacefully between its four familiar walls. A collection of textile samples lay spread out on the table – Samsa was a travelling salesman – and above it there hung a picture that he had recently cut out of an illustrated magazine and housed in a nice, gilded frame. It showed a lady fitted out with a fur hat and fur boa who sat upright, raising a heavy fur muff that covered the whole of her lower arm towards the viewer. Gregor then turned to look out the window at the dull weather.
He decided it was far more interesting to stay where he was and see where the day was to take him.
PART 1 — The Honest Assessment
- AI tools can be wildly useful
- Consumer AI is still basic — doesn’t match the hype
- The technology is genuinely amazing — with a very big BUT
PART 2 — What’s Actually Broken
- Venture capitalism as the root cause: the gold rush, the Ponzi logic
- The “wild west” — tech bro capitalism running roughshod over ethics, democracy, society
- AI as a form of colonisation
- The coming crash
- Radical acceptance and lack of discernment as a compounding problem
- Agents have no friction of reputational risk — the physical world doesn’t work like software
- The OpenClaw fiasco as case study
PART 3 — What AI Actually Is (and Isn’t)
- AI is not magic
- AI doesn’t give you ideas — it makes ideas actionable
- It requires context, judgement, and domain expertise
- It can be unreliable and untrustworthy
- Good inputs = good outputs / garbage in, garbage out
- The two-way door policy: why software mistakes are survivable; why this doesn’t transfer to the physical world
PART 4 — The Autonomy Question
- What we hand over, and to whom
- Our own autonomy is in question
- The low executive function use case — useful but at what cost?
- Frankl: “In the past, nothing was possible but everything was meaningful…”
- Conscious choice as the answer
PART 5 — How to Actually Use It
- Stop asking “will it take my job” — ask “how can I use it to do my job better”
- Use it professionally, for specific tasks, with appropriate constraints
- Check everything — judgement belongs to people
- Core AI competency: Can you define useful work for your agent in a prioritised and efficient manner?
- Spot recurring needs first, then find the tool
PART 6 — Software Engineering Principles Apply
- Agentic workflows in closed systems: boring and repeatable is the goal
- Permanent record of everything done
- Everything gets checked
- Show a preview of what done looks like
- Time windows and repair plans
- Build for restart, not perfection
- Build one workflow, then add modules — core loop first
- Build for maintainability over cleverness
- Treat prompts like APIs, not creative writing
- Separate memory from compute and interface (portable and swappable)
PART 7 — Guiding Principles
- Reduce the human’s job to one reliable behaviour
- Always build a trust mechanism; default to safe behaviour
- Make outputs small, frequent, and actionable
- Prefer routing over organising
- Build design for restart not perfection
PART 8 — The Human Skills That Now Matter Most
- Management skills and domain knowledge: task decomposition, feedback, knowing what done looks like, iterative refinement, QA
- Tool overwhelm: spot recurring needs first
- Death by prompts: use text expanders, embed prompts in workflows
- Update suffocation: be ruthlessly selective about what you learn
- The fear of doing it wrong
- Context assembly quality
- Frontier recognition — share your failures
PART 9 — Knowledge, Apprenticeship, and PKM
- Apprenticeship and succession: people build knowledge through access to you — same dynamic applies to AI
- Focus on the system, not the app
- Use frameworks for thinking to produce unique connections
- Distill actionable insights
- The 5-questions prompt technique
PART 10 — Digital Hygiene and Autonomy in Practice
- Digital hygiene vs. digital empowerment
- The comfort trap: “We are all too damned comfortable”
- Barriers: cost, ease, knowledge
- Core tenets: responsibility, discernment, the right to decide
- Close old accounts, export your data
- Internet filtering for children
- Moving toward decomputation (offline workflows)
- Techsocialism as the alternative framework
CLOSING THOUGHT
- “The more you create, the more powerful you become…” — James Clear
- Design for your brain: your quirks, strengths, energy, passions
- Business is the practice of delivering value to people
- 2026: get over the hype, start seeing results — it will be about protocols and processes
