1983 · Quicksilva
Every byte of the Spectrum's 48 KB of RAM, one pixel each, 384 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 | Not audited. No audit has checked every comment, and the last sample of 111 comments had 17 wrong, so between 10 and 23 % of the listing’s comments are likely wrong. How listings are checked |
| Build analysed | Original English tape supplied by contributor |
| Copy | agent-draft |
| Tools | emulator: release ZEsarUX-13.0, ZEsarUX_linux-13.0-ubuntu24_x86_64.tar.gz, disassembler: SkoolKit 10.1 (PyPI), host: Linux x86_64 (managed cloud), checker: Independent gpt-6-astra sample checker, browser: Chromium 156.0.8078.79 (Debian system executable), Playwright Core via CDP, python: 3.12.14, node: 24.22.1 |
| Model | gpt-6.1-sol |
| Kit version | 0.0.312 |
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, claude-opus-5-5 | 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 325 comments. | |
| Sample 1 | gpt-6-astra | 17 of 111 comments wrong, so between 10 and 23 % of the listing is likely wrong. All 17 were then corrected. Stratified sample: padding 0/20, map 0/16, shapes 15/15 (premature page-viewer claim), written 2/60; population-weighted rate 6.6%. Corrected all 17 findings; full audit recommended. | |
| Port check | claude-opus-5-5 | Checked a page’s port or widget against the game and corrected 0 comments it showed wrong. Quality pass of the pages: the game's messages and screens captured on the kit's 48K machine, the follow, SCAN and ant step rules run live there, the page's facing rule checked against $8A91, every picture drawn from the listing's bytes, and the ant speed counters read from BASIC lines 140-190 and counted live; no comment needed correcting |
likely wrong the sample’s range corrected right, or likely right not yet measured
Sources read 6 November 2026, with online research authorized:
| Feature | Status | Evidence |
|---|---|---|
| Choose boy or girl and rescue the other character | confirmed | BASIC lines 610/620 swap labels and shape bases; B reaches the welcome and play screens; $8F80 recognizes the partner |
| Four camera viewpoints | confirmed | $8400; live 0/P/Enter/Space checks select 3/2/1/0; 0+Space selects 0 |
| V forward, M/Symbol Shift rotate | confirmed | $8060, $8900; individual-key and no-input controls; both turn keys together preserve facing |
| C jumps; forward+jump climbs | traced | Input flags at $8060, occupied-cell paths at $8800/$88D0; C flag tested live, but a complete legal climb route through the city was not recorded |
| Four grenade distances | confirmed | $8D00 gives fuse values 2/4/8/16; live launch checks and 36-pass keyboard routes. Collisions can shorten travel |
| 1 returns to the gate | confirmed | Engine $8009 returns to BASIC; line 70 runs retry initialization. It is a retry sequence rather than an instantaneous coordinate change |
| Scanner guides the rescue | confirmed | $8F80/$87D0; initial panel observed; distance and same-height recognition tested by controlled live calls |
| Ants chase and bite | traced | Five records updated by $80D3, pursuit $8A00, actor collision and bite event $88A0, health handling $8F00; a naturally reached bite was not recorded |
| Jumping on an ant paralyses it | traced | $8BD1 writes delay $FF; $8A5D skips that ant. A keyboard-only landing route onto an ant was not recorded |
| Twenty grenades, limited human health, dangerous explosions | confirmed | BASIC DATA initialization; ammunition 20 to 19 on every tested launch; $8D00, $87A0, $8F00 establish human/ant blast and health paths |
| Ten rescues and changing target positions | traced | BASIC lines 90/95/800 and all 37 DATA triples; controlled escape transition observed, but the ten rescues were not played end-to-end |
| Score adds remaining time times rescued count | traced | BASIC line 90 at $5E49, decoded directly; score widget tested against that expression. No full ten-rescue scoring run was recorded |
| Scrolling block city and occlusion | confirmed | $83B0, $84A0, $8500; play and camera changes observed; 512 compositor indices and full render checked independently |
| Title, timeout, rescue, medal, and effect sounds | confirmed | All 17 audible recordings regenerate from the listing; title/timeout checked against emulator port writes; all recordings render through the sound player |
| High-score registration mentioned by Wikipedia | open | Every BASIC line and the engine/message allocation were read. This image has a cumulative score display and restart, but no registration path was identified. The general background claim may describe another build; no missing feature is asserted |
facts.md.The gameplay consumer of the retained shape group $BC00–$BCFF is unestablished. The listing keeps and the Graphics tab displays those records. High-score registration remains a source discrepancy, with the search scope recorded above.
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.