i keep coming back to old pcs and machines i can take apart
i write c rust and assembly and build compilers boot code and windows tools
some of my boards are still waiting for their first power on
if i cannot tell what a program is doing i will write a smaller one
projects · core stack · bench notes · experience · dbyte site
every project here has a public repo some are useful tools and some are still experiments nothing is polished up to hide what is unfinished
DBYTE — my programming language with a lexer parser type checker interpreter bytecode compiler vm and native code work there is a freestanding kernel path too and the public repo is only a snapshot of what has been built
BONEBOX-01 — a small x86 real mode system that boots from its own sector and loads a shell from disk it has a filesystem and executable format with 99 passing host tests and an interactive qemu boot
A780L3B BIOS research — reading and pulling apart firmware from a BIOSTAR motherboard rom dumps chip reads with a CH341A disassembly patches hashes and recovery notes actual bytes from the chip matter more than a patch that only looks right
the other 20 projects
LOOM — an x86-64 os experiment where cells replace one another and the bootstrap gives up control some stages have machine traces and some are still unresolved this is not a finished operating system
SHINESEA — a rust filesystem image core built around contiguous extents mirrored catalogs checksums and recovery rules tests cover damaged images as well as successful reads not claiming a finished release from the work in the source tree
CINDER-16 — a 16 bit virtual computer written in io with fixed width instructions cycle counting strict image loading and a debugger that can step backward i wanted a cpu i could inspect from end to end
DBLYATEM — c experiments with pe images amd64 decoding and strict binary checks on linux low level groundwork rather than a complete windows compatibility layer
DBYTE-HOLYC-OPT — studying the original HolyC optimizer and trying more transformations around its intermediate code passes the compiler and native backend are still HolyC and the extra work is experimental
DBYTEPAD — a native win32 text editor written in c with file handling search replace diffs and optional DByte integration there are tiny executable experiments too but the smallest builds do not have every editor feature
DEADFLASH — a windows disk image writer that makes the target explicit and checks it again before writing flush readback and verification are part of the job the repo records seven passing ctest checks for the qualified build
DEADBYTE-RECLAIM — plain win32 software inventory removal and rollback the graphical interface passes an operation plan to the cli instead of inventing another privileged path
DBNET — windows networking diagnostics with throughput probes and controlled configuration changes package sources are tracked and external measurement tests require consent
DEADWIRE-HTTPD — a small http 1 point 0 static server with assembly oriented backends more attention goes to request bytes bad paths and measured hot paths than feature count
BYTE-KEEPER — m0te is a little digital creature running in a native rust program it has persistent state a room and a taskbar companion weird idea but there is working code and there are tests
DEADGL — a c99 software renderer that takes text commands and produces framebuffer pixels ppm output and deterministic hashes i want to see what happens between the command and the pixel
DGP-1 — old style fixed function graphics hardware research with a command fifo triangle coverage texture units scanout memory arbitration rtl and software reference models no finished physical graphics card
dbyte-dbpx — a small lossless raster image format with explicit chunks and pixel formats built so the decoder can be read without dragging in a massive image library
DBYTE-370 — a socket 370 atx motherboard design for a pentium iii and via chipsets cpu and chipset pins sdram agp power clocks and reset are the work here not claiming a board has been powered on
S3CORE01 — a small rv64 computer design around the sg2002 one four layer cad revision has zero drc violations and zero unconnected items that does not mean a physical board or linux boot has been proven
USBDEADFLASH — a usb flash drive designed from the connector inward controller nand controlled impedance power and kicad layout still a design and not a manufactured drive
RIGOS — an old usb bootable cpu mining appliance experiment kept here for the record i stopped working on this one and do not maintain it anymore
JOBSIM-DESKTOP — keyboard and mouse compatibility work for job simulator archived as an engineering alpha rather than passed off as a finished release
Shower Game remaster study — source provenance and clean room planning for an old game research only and not a playable remake or a copy of someone else's files
languages systems tools and hardware — 37 badges
i do not use all of this every day this is the stuff i have actually worked with
languages
systems
build and debugging
hardware
windows linux bios firmware software and hardware
most of this started with breaking things on my own machines and figuring out why bios setup and recovery rom reads with a CH341A memory and storage trouble windows internals linux builds and boot failures
on the software side i have worked with parsers linkers executable formats win32 controls disk access sockets bytecode virtual machines x86 boot stages interrupts fixed point graphics and native tests i use debuggers disassemblers qemu compiler warnings regression tests and raw binary comparisons checking the bytes is the part that never changes
on the hardware side i use kicad for schematics footprints buses layer stacks power nets and drc i check pin maps against documentation and work on old chipsets firmware small stepper motors and salvaged hardware cad and simulation are not proof that a real board works
i also make game mods and investigate runtimes including my summer car reading someone else's code keeping the original behavior intact and tracing a value back to its actual source are all part of that work
this is independent engineering and personal research no made up employers no invented job titles and no pretending a prototype has shipped
what counts as proof
source identify the exact revision
build record what compiled
run keep the output and failures
compare check bytes and hashes and behavior
hardware measure the real thing before claiming it works
a passing unit test does not prove a pcb works a benchmark with different conditions proves nothing a screenshot is not the program that made it
courses and certificates
| coursework | issuer | record |
|---|---|---|
| MIPS Computer Architecture and Performance Optimization | BITS Pilani and Coursera | course certificate |
| Google IT Support | Google and Coursera | professional certificate |
| IBM IT Support | IBM and Coursera | professional certificate |
| CompTIA A+ Core 1 coursework | Total Seminars and Coursera | coursework specialization not CompTIA certification |
| Computer and ICT Literacy Skills, 45 Hours | Khon Kaen University and Thai MOOC | certificate of completion |
i do not take these numbers seriously open the code if you want to know what works
the rest is in the public repositories the code and the failures are there too



