1985 · Microdeal
Every byte of the Commodore 16's 16 KB of RAM, one pixel each, 128 to a row. Lit pixels are what this game is: its code, graphics, level data, sound, text, tables and variables, exactly where they sit in memory. Dim areas are screen and working memory; dark is unused.
| Tier | silver |
|---|---|
| Coverage | 100 % of the bytes the game uses have a description |
| Comments checked | Audited & low error rate. Every comment was audited once, then the last sample of 80 comments had 2 wrong, so between 1 and 9 % of the listing’s comments are likely wrong. How listings are checked |
| Media | The program file holds 1 file. Of those, 1 is in the listings. |
| Build analysed | Contributor-supplied packed PRG, SYS4109; engine entry $1100; original release identity unknown |
| Copy | agent-draft |
| Tools | emulator: plus4emu 1.2.11 (git 939e652) with kit/c16/codemap.patch, built without libsndfile, in the kit's server, disassembler: regenerator2000 0.9.20, python: 3.12.3, host: Linux (Ubuntu 24.04.3), x86_64 |
| Model | gpt-6.1-sol |
| Kit version | 0.0.323 |
Every step that measured the Source tab’s comments or made them more accurate, in the order it happened, with the models that did it. Accuracy of the Source explains each step and what the row above makes of them.
| Step | By | What it found | Of the listing |
|---|---|---|---|
| Research | gpt-6.1-sol | Looked the game up online (its manual, reviews or wiki pages) before reading its code. | |
| Annotation | gpt-6.1-sol | Wrote the listing’s comments. 100 % of the program is described, in 249 comments. | |
| Sample 1 | gpt-6-astra | 16 of 80 comments wrong, so between 13 and 30 % of the listing is likely wrong. All 16 were then corrected. 60 written and 20 generated comments; weighted rate 13.2%; one additional unverifiable purpose is recorded as unknown | |
| Audit 1 | gpt-6-astra | Checked all 249 comments and corrected 10. and 44 document/widget groups; 12 factual/widget findings corrected; footprint reverified | |
| Sample 2 | gpt-6-astra | 2 of 80 comments wrong, so between 1 and 9 % of the listing is likely wrong. Both were then corrected. 60 written and all 20 generated comments; weighted rate 0.8%; two homogeneous-pickup order caveats corrected |
likely wrong the sample’s range corrected right, or likely right not yet measured
reference/web-cover-back.gif. Primary controls and rules source.reference/.orientation.md.The Dragon/CoCo release and Game Boy remake describe other programs. Their controls and rules are not attributed to this C16 image.
| Feature | Status | Evidence |
|---|---|---|
| Collect every object, then leave through the top exit. | confirmed | draw_maze_packet counts object packets; collect_object decrements the count; move_player requires zero objects and player_y=0 for the upward exit. Original-code comparisons exercise all ten object sets. |
| Guards move through the corridors and kill Cuthbert on contact. | confirmed | Enemy update and direction selection at $1C47/$1DCC/$1E64; player contact range $69-$F9 at $1A34; live first-maze references. |
| Keyboard movement is RUN/STOP up, COMMODORE down, L left and colon right. | confirmed | Row/mask tables $136D/$1372 and controlled live input flags from the same boundary. Allowed down/right moves and blocked initial left/up moves are observed. |
| Joystick port 1 provides movement and fire. | traced | $1378 selects $FB and reads direction bits plus bit 6 for fire. The emulator's joystick hardware checks passed; the game-specific control probes used keyboard equivalents. |
| Starting Cuthberts. | differs | The cassette says three. This image shows spare digit 4 and reaches game over after five live forced timeouts. |
| Objects increase the score. | confirmed | Key/ring/apple parameters are 1/2/3 hundreds. Every original object record is tested through the collector. The adder loses overflow remainder, so actual gains can differ from face values. |
| Finishing quickly earns a time bonus. | confirmed | Reset shows 20 segments; one is consumed per 100 main-loop updates. Completion makes four ten-point additions per remaining segment. This clock follows game updates, not fixed real seconds. |
| Extra lives come from score milestones. | confirmed | $1952 compares the ten-thousands character before redraw. Original-code tests exercise the limit error and every award through its zero-byte failure. |
| Three starting bombs. | live | $7B=3 at the verified first-play boundary; a legal bomb route ends with two. |
| Fire lays a bomb and a second press detonates it. | differs | The code accepts a held fire flag after a six-update delay. Holding SPACE from first play lays and detonates the bomb without releasing it. |
| Explosions can kill Cuthbert and guards. | confirmed | Player range check and guard checks accept fire glyphs $69-$6C. A guard-kill test confirms the score call; the legal held-fire route produces five live explosion deaths. |
| An extra bomb is awarded for a completed screen. | live | Forced completion of level 255 advances to 0 and changes bombs 3→4 through the original transition. |
| Up to nine bombs can be stored while five icons are drawn. | confirmed | Completion caps $7B at nine; $1F49 clamps drawing to five. A live inventory of nine draws five icons. |
| Ten layouts repeat. | confirmed | Ten 24-byte records at $2011-$2100. Decoder matches full original characters/attributes for all 256 level numbers; all ten layouts have live screenshots. |
| Guard count depends on the tens digit. | confirmed | Four for tens=0, five for tens=1, six for tens=2-9. Original-code tests include all level bytes and verify the active slots and spawn addresses. |
| Enemy shapes and colours repeat every four/eight levels. | confirmed | Eight-entry tables at $1B97/$1B9F; every level byte is compared with the original initialiser. |
| Level 255 is followed by 0. | live | Original transition with the normal exit preconditions produces level byte 0 and returns to the next gameplay loop. The boundary is forced, not reached by a full played route. |
| Extra-life limit has a 208-attempt boundary. | confirmed | $FF permits 208 decrements to game over. The next award wraps the inline character to zero and enters BRK at $1A27, verified live in the C16 monitor. |
| Game over retains a high score and shows mazes until fire restarts. | traced | $118D compares/copies the six digits, clears the score, chooses timer-derived nonzero levels and polls fire. Live death checks stop at its entry; no claim of a separate title screen is made. |
| P pauses and R resumes. | live | Matrix polling at $1339; P stops in the scan loop and R returns play to $1172. |
| Graphics and sound. | confirmed | Original-data frame reconstruction matches 64,000 pixels; all source frames are shown on Graphics. Three audible sound excerpts regenerate from original-code input routes. |
The symbol map for this game is symbols.json; the listing behind the Source tab is listing.json. Write-ups, facts and symbol maps are CC BY-SA 4.0. The game itself is not hosted here.