Roast My Family
Get a loving, suspiciously accurate roast of everyone at home.
FIND YOUR NEXT THING TO TRY
Fresh prompts, better versions and experiments we just finished testing.
49 things to try
Get a loving, suspiciously accurate roast of everyone at home.
A ready-to-play quiz built around your family’s favourite things.
It produces a complete, age-aware game with answer key, bonus rounds and almost no setup.
Turn a chat with a grandparent into stories worth keeping.
The warm sequence and sensitive follow-ups create genuinely meaningful conversations worth keeping.
A bedtime adventure starring your child and their oddly specific powers.
Strong personalisation, safety limits and a calm ending make it reliably useful at bedtime.
A no-shopping game using whatever you already have.
It respects the players, time and objects you actually have, so the game is easy to start.
Face an AI opponent in an escalating battle of terrible jokes.
The turn-by-turn scoring creates a fun interaction, though the idea has limited depth.
Pick a ridiculous side and defend it far too seriously.
Clear rounds and playful guardrails make a good tiny game, but it is deliberately niche.
A spectacularly bad film pitch with a straight face.
The detailed format generates consistently funny results, even if it is entertainment only.
Let AI make a playful guess from five ordinary choices.
It is honest about being entertainment, but broad personality guesses can still feel generic.
Explain your cat’s suspicious behaviour with unnecessary evidence.
The strict documentary tone and fictional disclaimer make a simple joke land reliably.
Your ordinary week, narrated like prestige television.
A strong trailer structure lifts ordinary material, but the result depends heavily on your input.
Turn the random things in your kitchen into an actual dinner.
It combines real constraints, missing ingredients and food safety into an immediately usable answer.
Seven realistic dinners with one sensible shopping list.
Ingredient reuse and one grouped shopping list solve the whole planning problem, not just tonight’s meal.
Rescue cooked leftovers without pretending they are something they’re not.
Putting storage and reheating checks first makes this both practical and unusually responsible.
Adapt a meal without making adults eat beige food.
The shared-base and split-point approach gives families one realistic dinner instead of three.
A relaxed little trip planned around how you actually travel.
It plans around distance, energy and verification, though live opening times still need checking.
Find the overrated stops and better alternatives nearby.
Fair counterpoints and nearby alternatives are useful, but destination advice can become subjective.
A family day where the children are participants, not luggage.
It treats attention, toilets, snacks and downtime as real constraints, which makes the plan usable.
Compare cheap, comfortable and treat-yourself versions side by side.
The side-by-side budgets reveal useful trade-offs, but current prices must be verified elsewhere.
Clear, confident email that still sounds like you.
It preserves facts and voice while making the request clearer—simple, repeatable and genuinely useful.
Walk in knowing the goal, questions and likely awkward bits.
Goals, objections, questions and decisions turn vague context into a compact preparation brief.
Get a useful critic, not a cheerleader.
It forces assumptions, counterarguments and failure conditions into view instead of merely agreeing.
Extract decisions, owners and loose ends from messy notes.
It reliably converts messy notes into owners, deadlines and decisions without inventing missing facts.
Turn a confusing table into plain-English findings and questions.
It separates what the data shows from what still needs context, reducing confident misreadings.
A clear explanation that is simple without being childish.
Plain language plus a comprehension check teaches without simply replacing detail with baby talk.
A patient lesson that adapts one question at a time.
The adaptive questioning is strong, although the quality depends on sustained back-and-forth.
Practice that targets the gaps instead of repeating what you know.
Repeated retrieval and targeted correction make it a solid learning loop rather than a one-off quiz.
Find three different mental pictures for one difficult idea.
Multiple analogies make difficult ideas accessible, but every analogy has limits that need stating.
A low-pressure role-play that corrects you without stopping the conversation.
Short role-play turns and compact corrections create useful pressure without breaking the conversation.
A ten-minute mystery with clues, suspects and a fair solution.
The interactive structure creates a complete small game, though it needs a few minutes of attention.
A thoughtful conversation with the person you might become.
It prompts useful reflection while staying clearly speculative rather than pretending to predict you.
Turn everyday habits into heroic skills and questionable stats.
Shared details become playable characters quickly, with strong fun value but limited wider use.
David-Attenborough energy for your least majestic moments.
The precise voice and structure make ordinary details funny, even if the value is purely creative.
A polished launch for a product nobody needed.
It gives consistently inventive satire, but the joke has less replay value than our top prompts.
Adapt one useful thought without copy-pasting it everywhere.
It genuinely adapts structure to each platform and explicitly blocks fake stories and invented data.
Use your sample as a style guide, not a photocopier.
It uses observable traits instead of imitation, preserving voice while keeping the wording original.
A reusable image prompt for one character across a whole story.
The character sheet and reusable lock paragraph solve consistency across later image generations.
Can AI really read your personality from a playlist? We tried it.
It can describe playlist patterns, but the personality conclusions remain mostly projection.
Can a photo of your fridge produce dinner? Mostly, with one important check.
The confirmation step catches visual mistakes, making a clever trick useful rather than overconfident.
A good planner is not automatically a live price engine.
It can improve your search strategy, but it cannot guarantee the cheapest live fare without current data.
Turn a Garmin export into a cautious, evidence-based training review.
It checks the available fields, separates evidence from inference and builds a conservative plan without hiding the paid connector requirement.
A cautious starter plan built around your real life, not a fitness fantasy.
It gathers the right context, protects against unsafe personalisation and produces a conservative plan that fits the user’s actual time.
A six-week routine connecting training, movement, food habits and recovery.
The phased structure connects training, movement, food habits and recovery, though its broad scope still requires honest user input.
A bodyweight plan with real progressions, regressions and time limits.
Clear effort targets, regressions and one-variable-at-a-time progression make this unusually complete for a bodyweight prompt.
A realistic routine focused on habits and trends—not crash-diet promises.
It rejects crash-diet promises and focuses on repeatable habits, but individual weight response remains too complex for an AI to predict.
A progressive home strength plan built around your equipment and recovery.
Progression, reps-in-reserve and repeat-week rules turn limited equipment into a coherent strength plan rather than a random exercise list.
Flexible meals, substitutions and one grocery list—without fake precision.
Budget, location, substitutions and flexible portions create a genuinely usable plan while keeping medical nutrition outside the AI’s role.
A five-agent research system that looks for evidence before it falls in love with an idea.
Its evidence labels, sceptical validator and file-backed memory make research unusually traceable; the five-agent setup is still heavier than most people need.
Let AI challenge the purchase before it recommends another shiny thing.
It starts with your real constraints, separates needs from wants and demands dated evidence before recommending a purchase.
Specific details and a clear kindness rule make the result personal without becoming mean.