Considered Fluency · self-hosted · self-tracked

My AI Fluency Plan

By the gate date, I can delegate, describe, discern, and be diligent with AI on my real work: I choose the right tasks, specify them precisely, judge the output, and own the result.

Default plan: 35 sessions · ~35 hours · six sprints. This file is yours: set it up in the panel below, no code needed. Progress is stored only in your browser.

Make this yours — no code required

Fill in the three fields above and save. Title, objective, and gate apply immediately and are remembered by your browser; the file itself is untouched.
Work the cadence. Tick sessions as you complete them; review the Measures table weekly; keep the pace counter honest.
Optional, for the comfortable: the plan is plain HTML. Open it in a text editor to prune or add sessions (each is a small block with a unique data-id and an data-h hours value); counters recalculate automatically.

Where does my data go?

Nowhere. This page makes zero network requests. Completion state lives in your browser's local storage under the key named on the <body> tag. Clearing site data clears progress, so use the export button to keep a copy. Moving to a new machine: export, then keep the JSON with the file.

Complete

Hours done

0 / 0

Sessions

Days to gate

–

Pace needed

– h/wk

Measures

Feelings do not pass gates; measures do. Review this table weekly. Replace the targets with your own; keep every figure honest about how it was obtained.

#	Measure	Target at gate	How measured	Cadence
M1	Sessions completed	100% of the plan you kept after pruning	The counter above	Weekly
M2	Weekly hours vs required pace	At or above the pace counter	Hours ticked ÷ weeks elapsed	Weekly
M3	Competency self-scores (4D)	+1 or better on each, with one line of evidence per score	Baseline in Sprint 1 vs re-assessment in Sprint 6	Sprint 1 and 6
M4	Shipped artefacts	One real deliverable produced with AI assistance, plus your one-page personal use policy	Artefacts exist and are dated	Sprints 4–6

Six sprints

The plan is structured on the four AI fluency competencies — Delegation (deciding whether, when and how to engage AI), Description (specifying goals precisely), Discernment (judging output quality), and Diligence (owning the result) — then applies them to one real project and closes with evidence. Sessions bind to competencies, not to any vendor's tool, so the plan survives tool churn.

S1Orient and delegate 0% · what to hand over, and why

You cannot delegate work you have not mapped. This sprint produces your task inventory, your baseline, and your gate.

Day 1 — Map the terrain

1.0h Map your work — List ten recurring tasks. Classify each: automate, augment, or avoid. This inventory drives everything after it. 1.0h First contact, done properly — One 30-minute working session with a frontier assistant on a real task from your inventory. Log what surprised you, both directions.

Day 2 — Just enough mechanics

1.5h How these systems work, just enough — Context windows, training vs inference, why models fabricate. Enough to predict failure, no more. 1.0h Delegation decision drill — Pick five inventory tasks. For each, choose the mode: automation (AI executes), augmentation (thinking partner), or agency (AI acts ahead). One written sentence of justification each.

Day 3 — Baseline and gate

1.0h Baseline self-assessment — Rate yourself 1–5 on Delegation, Description, Discernment, Diligence. One line of evidence per score. Date it. This is M3's denominator. 0.5h Set your gate — Rewrite the objective sentence in the header. Set data-gate. A plan without a date is a wish.

S2Describe 0% · specification is the skill

The quality of what you get is bounded by the precision of what you asked for. This sprint turns prompting from improvisation into specification.

Day 1 — Anatomy

1.0h Anatomy of a strong request — Role, context, task, format, constraints, examples. Take three of your old prompts and rebuild them on this frame; compare outputs. 1.0h Context is the work — Run the same task cold, then with source material supplied. Measure the difference; internalize why supplying context beats asking better questions.

Day 2 — Reuse

1.0h Iteration drill — One deliverable, three refinement turns, each turn naming a specific defect to fix. Stop when marginal improvement drops. 1.5h Templates for your top tasks — Build three reusable request templates for the highest-value tasks in your Sprint 1 inventory. These are assets; store them where you work.

Day 3 — Contracts

1.0h Format contracts — Practise demanding structure: tables, checklists, schemas, word limits. Output you can use without reformatting is the standard. 0.5h Describe review — Re-run Sprint 1's first-contact task with your new technique. Diff the two results; note what specification bought you.

S3Discern 0% · judge the output like a reviewer

Fluent users are not the ones who trust the machine; they are the ones who know exactly when not to. This sprint builds your verification habits.

Day 1 — Verification

1.0h Verification habits — Every load-bearing claim, number, citation, and name gets checked against a source. Practise on one AI-produced document; count what you catch. 1.0h Failure hunt — Deliberately elicit confident errors in your own domain, where you can spot them. Knowing the failure texture is the skill.

Day 2 — Rubric

1.0h Write your quality rubric — Five criteria for acceptable AI output in your main task: accuracy, completeness, tone, structure, and the one specific to your field. 1.0h Blind comparison — Score an AI draft and your own past work against the rubric, blind if you can arrange it. The result calibrates both your trust and your self-assessment.

Day 3 — Limits

1.0h Map the limits — Document where the tools fail in your domain: task types, data types, reasoning depth. This map is what makes your Delegation choices defensible. 0.5h Second-opinion pattern — Use one model to critique another's output against your rubric. Adversarial review is cheap; use it.

S4Be diligent 0% · you own the result

Whatever the tool did, the work carries your name. This sprint writes down the rules you will actually follow.

Day 1 — Data and disclosure

1.0h Data discipline — Write the list of what you never paste into a model: client identifiers, credentials, personal data, anything under confidentiality. If your organization has a policy, read it; if not, yours is now the policy. 1.0h Provenance and disclosure — Decide, in writing, when you disclose AI assistance and how you record what was machine-drafted. Ambiguity here is where reputations go.

Day 2 — Risk and regulation

1.0h Bias and failure modes — Study the systematic failure classes: bias in and out, sycophancy, spurious confidence. Identify which two are most dangerous in your work. 1.0h Regulatory literacy — Know what your jurisdiction expects. In the EU, Article 4 of the AI Act obliges organizations to take AI-literacy measures for staff; elsewhere, expectations arrive through professional bodies and employers. Know where you stand.

Day 3 — Your operating rules

1.5h Write your one-page use policy — Three headings: allowed, gated (needs review before use), never. Date it, follow it, revise it at the gate. This is M4's second artefact.

S5Apply 0% · one real deliverable, shipped

Courses end; work ships. Everything from Sprints 1–4 converges on one real deliverable that someone other than you will use.

Day 1 — Select and start

0.5h Project selection — One real deliverable from your inventory, meaningful but boundable. Write the done-condition before you start. 1.5h Build session one — Work the project with your templates and format contracts. Log where the tool carried weight and where it dropped it.

Day 2 — Build and critique

1.5h Build session two — Push to a complete draft. Resist polishing; completeness first. 1.0h Critique pass — Apply your Sprint 3 rubric and the second-opinion pattern. Log every defect found; fix the load-bearing ones.

Day 3 — Ship

1.5h Ship it — Finish, verify, deliver to its real audience. This is M4's first artefact. 0.5h Retrospective — What did AI assistance change: hours, quality, confidence? Numbers where you have them, honesty where you do not.

S6Consolidate and evidence 0% · prove it, then plan the next climb

Fluency you cannot evidence is fluency you cannot claim. This sprint closes the loop against your baseline and decides what comes next.

Day 1 — Re-assess

1.0h Re-assessment against baseline — Re-score the four competencies with one line of evidence each. Compare to Sprint 1. M3 passes or it does not. 1.0h Close the weakest gap — One targeted session on your lowest-scoring competency, using the drills from its sprint.

Day 2 — Evidence

1.0h Assemble the evidence pack — Export your progress JSON, gather the dated artefacts: inventory, baseline and re-assessment, templates, rubric, limits map, use policy, shipped deliverable. 1.0h Teach one thing — Explain one technique to a colleague until they can use it. Teaching is the test you cannot fake.

Day 3 — Gate review

1.0h Draft the next cycle — Copy this file, reset the storage key, and write the next six sprints: deeper on one competency, or wider into new task classes. 0.5h Gate review — Score the Measures table. Pass, extend, or redesign; all three are legitimate, undocumented drift is not.

Risk register

Self-directed plans rarely fail on content; they fail on these. Each risk has a mitigation already wired into the plan.

#	Risk	Signal	Mitigation (already in the plan)
R1	Momentum fades after Sprint 2	Pace counter climbs above your available weekly hours	Weekly Measures review; "Open only" view keeps the remaining work small and visible
R2	Passive consumption replaces practice	Sessions ticked but no artefacts exist	Nearly every session produces a written artefact; M4 audits them
R3	Tool churn strands the plan	A favourite tool changes or disappears	Sessions bind to the four competencies, not to any vendor
R4	No feedback loop; errors compound	You cannot name the last mistake you caught	Rubric (S3), second-opinion pattern, teach-one-thing (S6)