Build time and release eligibility are different clocks.A playable build in 14 days is not a promise of a Steam launch. For a first release, Valve documents a 30-day waiting period after the application fee and at least two weeks of a public Coming Soon page. Store/build review and demo setup need their own schedule. Verify your account and application state before choosing a launch date.
Valve’s release requirements ↗ Desktop companionDemo · Focused
01Desk Lighthouse
A tiny lighthouse sits at the edge of your screen. Each completed focus session guides a different paper boat home.
The spin: Progress is an evolving harbor with finite hand-authored arrivals, rather than a second idle economy to manage.
Build exactly this
- One resizable window and optional always-on-top mode
- One timer, one save slot, six boats and three harbor upgrades
- One 25-minute ritual plus a fast preview mode
Cut from the sprint
- No calendar accounts, desktop injection or cloud sync
- No procedural dialogue, pets economy or multiplayer
- Where AI helps
- Scaffold timer state, save migrations and UI variants; generate deterministic test cases for pause, sleep and resume.
- Human bottleneck
- Comfort, idle CPU use, sound restraint and whether interruption feels pleasant.
- First clip
- An empty harbor becomes a lit miniature town while the real work stays visible.
- Where to market
- Cozy and productivity curators; a short desk setup clip linked to an honest playable sample.
- Proposed validation
- Five remote workers use it during one real focus session. Record voluntary second-session use and interruptions they disliked.
- Revise or stop if
- Stop if the app repeatedly steals attention or its appeal disappears when the novelty clip ends.
Risks and release scope
- OS-specific window behavior consumes the sprint
- Existing desktop bundles are strong competition
Treat the two-week output as a comfort test. A shippable desktop utility requires more compatibility coverage.
Read the demand and competition brief →Inventory strategyDemo · Stretch
02Lost Luggage League
Pack the wrong suitcases for the right passengers: every item changes the next automated baggage-carousel duel.
The spin: The opponent is visible before packing. Rearranging a small suitcase is the decision, not collecting hundreds of items.
Build exactly this
- One 4×4 grid, eight items, three enemies
- Five encounters with deterministic combat
- One visible synergy and one counter per item family
Cut from the sprint
- No online PvP or matchmaking
- No rarity economy, procedural map or meta-progression
- Where AI helps
- Implement grid placement tests, combat logs and a seeded balance simulator; propose item combinations for human review.
- Human bottleneck
- A readable and interesting choice must survive the very small content pool.
- First clip
- Rotate a ridiculous umbrella and turn a losing suitcase into a winning combination.
- Where to market
- Inventory and autobattler creators who explain decisions in long-form videos.
- Proposed validation
- Ask ten players to explain why they changed their layout before the third encounter; log distinct successful arrangements.
- Revise or stop if
- Stop if one obvious layout solves every encounter or players cannot explain a synergy.
Risks and release scope
- Content combinatorics grows quickly
- A strategy audience expects depth beyond a funny skin
Two weeks can test one ruleset, not compete with an established item ecosystem.
Read the demand and competition brief →Finite incrementalMicrogame candidate · Focused
03Bureau of Gravity
Shred obsolete gravity permits until the office furniture starts floating; buy upgrades that make the mess wonderfully worse.
The spin: Progress visibly transforms one room and ends with a controlled comic disaster.
Build exactly this
- One repeatable input and three upgrade branches
- One room with staged transformations
- 15–25 minutes of progression and a real ending
Cut from the sprint
- No prestige loop, live economy or infinite content
- No general physics simulation; use authored stages
- Where AI helps
- Generate UI scaffolding and offline progression simulations; draft tuning tables and edge-case tests.
- Human bottleneck
- Pacing, sound, feedback and avoiding a dead stretch between upgrades.
- First clip
- A dull desk becomes an orbiting stationery storm in a single before-and-after shot.
- Where to market
- Incremental-game communities where self-promotion is allowed and small creators who finish short games.
- Proposed validation
- Observe ten complete sessions. Locate the first boredom or confusion point; measure how many finish voluntarily.
- Revise or stop if
- Cut progression if an upgrade does not change what the player can see or decide.
Risks and release scope
- Fast content consumption and refund exposure
- A loop can be satisfying in a clip but tedious to play
Potentially a complete tiny game only with an already-ready Steam application, licensed assets and a stable build. Price after testing the finished experience.
Read the demand and competition brief →Tactile job simulationDemo · Focused
04Midnight Returns Desk
Process returns from customers who insist impossible products are working as intended.
The spin: Accepting an item changes the rules for later customers, so a familiar inspection job becomes comic cause and effect.
Build exactly this
- One counter, six products and eight customers
- Inspect, stamp and place as the only verbs
- One authored shift with two endings
Cut from the sprint
- No walkable town, staff management or stock economy
- No generated conversations or voice service dependency
- Where AI helps
- Build an authored customer-state machine, interaction prompts and localization drafts; test each rule branch.
- Human bottleneck
- Objects must feel tactile, and jokes need timing and legibility.
- First clip
- A customer returns a toaster because it keeps predicting the future; the player tests it.
- Where to market
- Simulation and comedy creators; show an actual shift with a complete setup and consequence.
- Proposed validation
- Have ten players complete a shift without explanation; record misunderstood rules and which objects they retest.
- Revise or stop if
- Simplify if the player reads more than they touch or the premise needs a long explanation.
Risks and release scope
- Animation can dominate production
- The premise may feel like a sketch without meaningful decisions
An eight-customer slice validates interaction; it does not validate a full shop economy.
Read the demand and competition brief →One-room horrorDemo · Focused
05Night Shift: Freight 13
Sort late-night parcels in a locked freight office; one parcel learns where you are whenever you make a sound.
The spin: The threat reacts to the same scanning and handling actions that earn your shift pay.
Build exactly this
- One room, eight parcels and one threat state machine
- One repeatable scanner interaction
- A 10-minute shift and one authored climax
Cut from the sprint
- No city, combat, skill tree or enemy navigation outside the room
- No endless anomaly catalog or online co-op
- Where AI helps
- Scaffold interactable states, subtitles and save/reset tests; generate debug views for threat transitions.
- Human bottleneck
- Spatial audio, anticipation, lighting and fair readable danger.
- First clip
- The parcel behind you repeats the sound of the scanner after you unplug it.
- Where to market
- Short-horror creators with a clean capture build and a short, spoiler-light pitch.
- Proposed validation
- Observe whether first-time players connect noise with the threat before the climax; ask where they felt cheated.
- Revise or stop if
- Remove random deaths and threats whose rule cannot be inferred from feedback.
Risks and release scope
- Audio and atmosphere require craft
- One-shot horror competes with abundant low-price releases
Build one satisfying shift before adding combat or a second location.
Read the demand and competition brief →Deduction puzzleMicrogame candidate · Focused
06Museum of Alibis
Reconstruct who stole a museum exhibit by matching three contradictory visitor recordings to a tiny floor plan.
The spin: Players alter the timeline and immediately see which alibis become impossible.
Build exactly this
- One case, four suspects, twelve clues
- A timeline and room map as the only screens
- A hint ladder and one final accusation
Cut from the sprint
- No voice acting, freeform AI questioning or procedural mysteries
- No branching narrative beyond the solved case
- Where AI helps
- Create clue bookkeeping tools and UI code; try to falsify the timeline with generated test scenarios.
- Human bottleneck
- The mystery must have one fair solution. A human verifies every dependency and clue.
- First clip
- Move an alibi by three minutes and reveal that two suspects cannot both be telling the truth.
- Where to market
- Puzzle and detective communities, then creators who enjoy explaining deductions.
- Proposed validation
- Watch eight blind solves. Track hint usage, false ambiguity and whether the final explanation matches the evidence.
- Revise or stop if
- Rewrite any clue that only makes sense after the author explains it.
Risks and release scope
- Writing and puzzle verification are the real workload
- One case may not meet audience expectations for paid content
A complete short case is possible; pricing and Steam scheduling remain separate decisions.
Read the demand and competition brief →Photo explorationDemo · Focused
07Pigeon Press Pass
Photograph pigeons committing tiny civic offenses in one compact square to finish a neighborhood newspaper.
The spin: The joke lives in the framing: the same bird can satisfy different assignments depending on what the photograph includes.
Build exactly this
- One courtyard, three birds and six photo assignments
- Simple subject and framing checks
- A printable-style front page as the final reward
Cut from the sprint
- No open world, complex animal AI or user-uploaded images
- No image recognition API; deterministic subject tagging
- Where AI helps
- Implement framing checks, photo UI and assignment validation; automate camera-boundary tests.
- Human bottleneck
- Authored behavior and charming compositions; consistent assets matter more than asset quantity.
- First clip
- A mundane bird becomes a scandalous newspaper front page through a perfectly timed photo.
- Where to market
- Cozy/photo-game creators and screenshot-led community posts with one playable square.
- Proposed validation
- Ask players to solve assignments without arrows; record whether they take extra photos after finishing.
- Revise or stop if
- Stop expanding if all assignments resolve from one camera angle.
Risks and release scope
- Charm is not a proven sales gap
- Reference games have very different review scale
Validate composition and playful observation before funding additional locations.
Read the demand and competition brief →Minimal logisticsDemo · Stretch
08One Night Courier
Route three nocturnal couriers through a small town where streets disappear when the last shop closes.
The spin: The route network shrinks on a schedule visible from the start; players plan around temporary access.
Build exactly this
- One graph of twelve junctions
- Three couriers, four package types and six authored scenarios
- A clear clock and instant replay
Cut from the sprint
- No traffic physics, city generator or multiplayer
- No dynamic economy or dozens of buildings
- Where AI helps
- Build the pathfinding sandbox and deterministic simulation tests; generate candidate levels for human validation.
- Human bottleneck
- Readable routing and scenario quality. Random maps are not automatically good puzzles.
- First clip
- The last bridge closes as three deliveries converge, then one clever reroute saves the night.
- Where to market
- Strategy/puzzle creators who show a problem and its solution; compare two short gameplay explanations.
- Proposed validation
- Give ten players the first three maps. Record avoidable confusion, restarts and meaningful route alternatives.
- Revise or stop if
- Cut any system that adds visual noise without changing a route decision.
Risks and release scope
- Simulation bugs become design confusion
- Successful reference games represent years of iteration
A deterministic scenario pack is the two-week target; endless mode needs a separate validation cycle.
Read the demand and competition brief →