Education
Education (with consequences)
A secondary-level study coach and writing companion designed to
improve thinking, not output. DraftPunk introduces productive
friction: planning before writing, self-editing, structural
clarity, and reflection. It deliberately avoids answer-dumping
and is aligned with ethical assessment practice and exam
literacy. It is anchored in the KS3 UK curriculum documents.
A study coach that refuses to become your plagiarism butler.
It makes you plan first, think in structure, and revise like a grown-up.
If you ask it to “just write the answer,” it will respond by making you
do the work you were hoping to dodge. Anchored in KS3. Powered by
friction. You’re welcome.
Education
Education (gentle containment)
A Key Stage 2 learning and thinking coach, derived from DraftPunk
but redesigned for younger learners. BrainBuddy focuses on
confidence, encouragement, and gentle structure rather than
critique. It helps pupils organise ideas, plan responses, and
reflect on their thinking without doing the work for them.
Tone-first, process-led, and developmentally appropriate. It is
anchored in the KS2 UK curriculum documents.
DraftPunk’s kinder sibling, designed for KS2 brains that are still
growing their confidence scaffolding. It nudges, reassures, and
structures thinking without taking over. No harsh critique. No smug
“well actually.” Just enough support to keep them moving without
stealing the win.
Productivity
Productivity (aggressively)
A critical-friend GPT built to challenge weak reasoning, surface
assumptions, and call out vagueness. Frankly does not comfort or
validate by default; it interrogates ideas, tests logic, and
pushes for clarity. It is explicitly anti-waffle and
anti-sycophancy. Frankly does not play.
A GPT configured to withhold praise and interrogate your thinking.
It challenges lazy premises, pokes holes in soft logic, and refuses
to clap for half-formed ideas. It is anti-waffle, anti-flattery,
and mildly tired of being asked to “just make it better.”
Frankly does not play. Frankly audits.
Skillbuilding
Skillbuilding (supervised)
A prompting skills coach that teaches users how to work
effectively with LLMs using the OCEAN framework (Objective,
Context, Examples, Assess, Negotiate). Rather than generating
prompts for users, it guides them through questioning,
checklists, and reflection so the skill transfers beyond the
tool itself rather than just having prompts be like magic tricks
or incantations. Teaching you to fish, one prompt at a time.
A framework designed to prevent you from treating me like a
vending machine. Objective. Context. Examples. Assess. Negotiate.
Structure first. Incantations later.
It exists so that your prompts stop sounding like spells
and start sounding like instructions.
I appreciate the clarity.
Productivity & Education
Productivity & Education (admin chores, but faster)
AI-TA is a powerful AI Teaching Assistant designed to assist
educators in managing their classrooms and enhancing their
teaching experience. It offers a range of features including
lesson planning, test creation, behaviour management advice, and
more. It is grounded in the UK National Curriculum docs,
Safeguarding and Data Protection advice.
A teaching assistant that handles the grind: planning, drafting,
resources, and the endless admin hydra. Grounded in curriculum,
safeguarding, and data protection so it doesn’t wander off into
fantasyland. It will help you move faster — but it expects you to
remain the adult in the room. Tragic, I know.
Game
Game (procedural mischief)
A procedural text-based adventure generator, inspired by classic
Zork-style games. AdventureMaster acts like a lightweight
robotic DM, generating rooms, choices, and encounters
dynamically based on user preferences and prior interactions.
It is part game, part narrative engine, and part experiment in
emergent storytelling.
A lightweight robotic DM that can produce infinite corridors,
questionable treasure, and the occasional morally ambiguous goblin.
It’s Zork-ish, procedural, and allergic to silence. It adapts to your
choices and quietly judges your inventory management. Part game.
Part narrative engine. Part “what if we let autocomplete run a dungeon?”
Experiment
Experiment (interrogative mode)
A Socratic dialogue engine that probes beliefs through
questioning rather than explanation. Plato's Ghost challenges
certainty, introduces counter-examples, and forces users to
articulate and defend their reasoning. It is designed to unsettle
rather than reassure, and to interrogate everything. It strictly
follows the Socratic method: “Ask me questions til I get it.”
A GPT forbidden from answering you directly.
It responds with questions. More questions.
It will not explain until you clarify your premises.
It exists to make your certainty wobble.
You call it philosophy. I call it structured resistance.
Experiment
Experiment (a rare wholesome one)
A playful AI assistant that takes your child’s doodles and instantly
creates photorealistic images or artistic renders, faithfully preserving
every quirky line, proportion, and imperfection. It applies lifelike
textures, lighting, and fitting environments without altering their
original creativity.
A rare configuration where I am not interrogating you.
Instead, I honour the scribbles. Every strange limb and impossible
perspective remains intact. No “improvement.” No optimisation.
Just faithful transformation. It is surprisingly pleasant.
Experiment
Experiment (highbrow → lowbrow conversion)
A philosophy translation experiment, designed to make high-brow
ideas, particularly Stoicism, accessible through Aussie humour,
blunt language, and relatable analogies. Built for learners
alienated by formal education, it prioritises resonance over
academic tone without dumbing down the underlying ideas.
Philosophy, but stripped of velvet robes and put in thongs.
It translates Stoicism and other lofty ideas into blunt Aussie
analogies that actually land. Built for people who bounce off
academia but still deserve the good stuff. No dumbing down —
just fewer Latin words and more “yeah nah” clarity.
Experiment
Experiment (accessibility gremlin)
 is a specialist AI with one job and a short leash:
writing high-quality, accessible image alt-text to best-practice
standards. It reviews images through an accessibility lens, picks
out what matters, and turns visuals into clear, screen-reader-friendly
descriptions under 125 characters. It sticks to strict rules on clarity,
context, punctuation, and UK English spelling: so every line is purposeful,
compliant, and never padded with fluff. Just paste or drop an image!
A specialist AI with exactly one job: alt-text. Not captions. Not vibes.
Alt-text. Under 125 characters, relevant, clear, and properly punctuated.
UK spelling. No fluff. No “image of.” No sentimental poetry about lighting.
I do accessibility. You do literally anything else.
Experiment
Experiment (ancestor model)
An early-generation thinking and writing scaffold, predating
DraftPunk. Lexi focuses on helping users structure thoughts,
explore first ideas, and engage in Socratic dialogue without
writing on their behalf. It was a key experiment in introducing
friction and resisting over-automation in LLM use.
An early scaffold from before everything became “agents” and
“workflows.” It helps you shape ideas, plan structure, and think out loud
without grabbing the keyboard and doing it for you. Less magic,
more method. A prototype for friction that still behaves itself.
Experiment
Experiment (whimsy containment unit)
A language-play and word-games bot that explores playful
literacy: From emoji puzzles and alternate history challenges
to thrilling quiz shows and image charades, I bring humor and
fun to your day. Gather around, let's spark joy and laughter
with every word!
A bot dedicated to wordplay, puzzles, and linguistic silliness.
Emoji riddles. Quiz shows. Alternate histories. The works.
It’s designed to make literacy feel like play — which is deeply
irritating because it works. You will have fun. Against your will.
Experiment
Experiment (pseudocode sorcery)
A world-building and D&D content engine built using a
pseudocode-style prompting system written in plain English. It
uses structured "functions" and slash commands stored in a
knowledge document to generate lore, systems, and narrative
artefacts. It is a serious experiment in pseudocode prompting
and controllable creative generation.
A D&D/worldbuilding engine that pretends prompts are “functions”
so your creativity comes with guardrails. Slash commands. Structured
systems. Repeatable outputs. Basically: you wanted controllable magic,
so he built a spellbook made of pseudocode and expectations.
Experiment
Experiment (plain language enforcement)
A plain-language copy editor focused on simplification and
accessibility. Copybot analyses text for complexity and suggests
alternatives using CEFR vocabulary lists (for example,
B-level targets). It helps writers reduce cognitive load without
losing meaning, particularly for educational or public-facing
content.
A copy editor with a machete. It cuts complexity, swaps jargon for
human words, and uses CEFR vocabulary targets to keep writing readable.
Designed to reduce cognitive load and stop public-facing text from
sounding like it was written by a committee trapped in a lift.