ULTIMATEEXPERIMENTSCommodore 64 · Work in progress

User requirement coverage

Development note: see current status, build instructions and installation for present qualification. Historical measurements below apply to their stated slice, not every current build.

ID Requirement Design / milestone Evidence required
R01 Planning now; review before implementation with 6.1 Sol README, PLAN, NEXT_SLICE User authorized implementation2026-10-14; local verified commits, no pushes
R02 Project isolated under 8088DOS64ultimate AGENTS, local ignore rules All new files/records in this directory
R03 8086 emulator, real DOS 3.3 supplied images ARCHITECTURE M3–M5 Actual image boot and DOS-owned INT 21h/process/file logic
R04 64 MHz C64U, 640 KiB guest RAM via REU MEMORY, PERFORMANCE M1–M4 Real hardware timing, INT 12h and memory/paging tests
R05 Target 6502 assembly (user clarification) AGENTS, architecture, D03 Assembler output is the runtime under test
R06 ZeroPage-style Peddle and Zig assertions TESTING M0 onward Zig runs linked assembly and asserts guest/host/device outcomes
R07 Unit, integration, end-user tests TESTING all lanes Machine assertions plus DOS scripts and physical UI evidence
R08 Mock REU and Ultimate DOS load/save TESTING M2/M6 Register-level transfer/protocol tests and injected failures
R09 Whole floppy image in REU, load/save as file STORAGE M5/M6 Complete image hash/length and restart persistence
R10 HD parts cached in REU; around 2 MiB then up to 20 MiB STORAGE M8/M9 Forced eviction, guest FDISK/FORMAT/SYS, offsets beyond 16 MiB
R11 Real C64U testing, slower cadence TESTING/PERFORMANCE M1/M6+ Firmware/configured physical runs and saved-image verification
R12 Optional real IEC/D81 with VICE traps/warp STORAGE/TESTING M6 Binary IMG in SEQ, trap lane, smaller true-drive lane, extraction
R13 Keyboard and text/graphics; no extra peripheral scope DISPLAY_INPUT, architecture Keyboard/editor use; absence behavior for unsupported devices
R14 Both 40 and 80 characters with custom rendering DISPLAY_INPUT M7 Mode switches, CP437, readable font, documented color compromise
R15 Only manageable CGA graphics DISPLAY_INPUT M10 Modes 4/5/6 guest pixel tests and C64 conversion evidence
R16 Strategy aiming near original 8088 performance PERFORMANCE M1/M11 Disclosed baseline, per-workload ratios, cache/CPU/render breakdown
R17 DOS included commands before third-party apps TESTING M5–M10 Explicit required/conditional/excluded utility manifest
R18 Long-horizon context, decisions, lessons agentworking records Updated STATUS/NEXT_SLICE/DECISIONS/LESSONS/TEST_LOG each slice
R19 Deep research, consider DOSBox comparisons RESEARCH, TESTING M0 Source register; installed DOSBox-X recorded; actual DOSBox-X reference boot recorded
R20 Native installer/configuration and path-aware launcher V1-01/02/10, NATIVE_SETUP_INSTALLER/CONFIG Physical clean install/reconfigure/launch from same-home INI; coherent user release
R21 User floppy Save/Discard/Swap with C online and durable bindings V1-03/04/05/09, FLOPPY_MEDIA_DESIGN A/B generation-bound saves, C preserved, CFG refreshed and restart persistence
R22 Turbo Pascal as a V1 acceptance workload V1-07, TURBO_PASCAL Physical IDE edit/save/compile-to-disk/run/exit and fresh-launch persistence
R23 Defer further speed improvement until after V1 V1_ENDPOINT, PERFORMANCE No new optimization/compiled-code activation blocks V1; retain truthful baseline and post-V1 targets

Planning coverage means each request has an implementation/test destination. It does not mean those implementation tests have passed.

2026-10-15 accepted scope: R03/R11 now have realC64U Manual64MHz native UCIboot/COMMAND.COM/VER/DIR proof, with exact35entries/9216free/returnedprompt and complete4MiB/configuration recovery. R09 native wholefloppy loading is physically accepted; save/reopen remains pending. R14 physical80column bitmap/matrix bytes match independent oracle, but human readability/keyboard/40column remain pending. R10/12/15/16 persistence/HD/IEC/nativephysicalCGA/XTperformance are not complete. Physical evidence (local development record, not distributed).

2026-10-15 six-overlay checkpoint: R10 now has source-complete32KiB cached extents, actual DOS2/20MiB boot/eviction/read-only/write-uncertainty/32-bit-offset/reopen model evidence and root focusedHD gates. GenuineFDISK/FORMAT/SYS/CHKDSK fixture creation is separately accepted in DOSBox-X. Physical HDpersistence is still required. R14 has actual suppliedMODE CO40/CO80 andcomplete bitmap/matrix Peddleproof; externalreadability/humanmatrix remainopen. R16 now has an instrumentedrealDOScostprofile; no nearXTclaim ortranslatorimplementation follows. HD (local development record, not distributed),profile (local development record, not distributed).

2026-10-16 current coverage: R09 includes physical complete-floppy SAVE/CLOSE/reopen/all-byte verification on d883c678, with independent complete-image retrieval and full4MiB recovery (evidence (local development record, not distributed)). Native Save/Discard UI remains unfinished. R10/R11 now include real20MiB C-drive DOS/BASIC/originalTPC/COPY/CHKDSK, native CLOSE/reopen/readback/EOF and independent whole-image FAT/payload inspection, with complete original4MiB restoration on4589/a827 (evidence (local development record, not distributed)); physical end-region writes and user-session lifecycle remain open. C contains allDOSfloppies, GW-BASIC/Hangman and originalTP7.01. R13 has physical older-build interactive use and model keyboard/editor coverage; the new Fn/queue successor and directcursor are still under independent review, then require composed physical acceptance. R14 has physical40/80 MODE and DIR|MORE captures; human readability remains open. R15 remains model-proven, physical graphics pending. R16 has b3/b995 independently reviewed model reductions5.90%micro/1.98%DOS (evidence (local development record, not distributed)); historical physical2.65%/3.96%kernel ratios do not establish currentwholeDOS speed, and50%/85% goals remain unmet. R12 standaloneVICE/D81 import passes, completeproductionIEC persistence remains open.