The editorial explicitly states 'nobody is going to fork 86-DOS and ship it' and frames the release's significance as historical rather than technical. The argument is that the 6,000 lines of 8086 assembly matter as preserved DNA of Microsoft's empire, recovered from SCP floppies via the Computer History Museum.
The editorial emphasizes that 0.1-C is 'a working operating system written by one person in roughly two months — kernel, shell, file system, and device drivers — and it fits comfortably inside a single browser tab.' This frames Tim Paterson's QDOS effort as a benchmark for what a single engineer can accomplish, contrasting implicitly with modern bloat.
By surfacing the Ars Technica coverage on HN with the headline framing 'the earliest DOS source code discovered to date,' the submitter positions this as a continuation of Microsoft's 2014 (MS-DOS 1.25) and 2018 (MS-DOS 2.0) releases. The framing treats this as part of an ongoing institutional preservation effort rather than a one-off curiosity.
On April 28, 2026, Microsoft's Open Source Programs Office published the source code for 86-DOS version 0.1-C to GitHub under the MIT license. The drop appeared on the [opensource.microsoft.com blog](https://opensource.microsoft.com/blog/2026/04/28/continuing-the-story-of-early-dos-development/) as a continuation of the company's 2014 and 2018 releases of MS-DOS 1.25 and 2.0, and it is — per Microsoft's own framing — the earliest DOS source code discovered to date.
86-DOS 0.1-C is Tim Paterson's pre-Microsoft code, written at Seattle Computer Products in 1980 as a 'Quick and Dirty Operating System' (QDOS) to give SCP's 8086 board something to run while Digital Research dragged its feet shipping CP/M-86. Microsoft bought the rights to 86-DOS in July 1981 for $75,000, rebadged it as MS-DOS, and licensed it to IBM as PC DOS 1.0 for the August 1981 launch of the IBM 5150. The 0.1-C source predates that acquisition. It is, in a literal sense, the DNA Microsoft built a forty-year empire on top of — recovered, the blog notes, from old SCP floppies surfaced by the Computer History Museum and cross-referenced against Paterson's own archives.
The repository is about 6,000 lines of 8086 assembly across a handful of files: `COMMAND.ASM`, `IOSYS.ASM`, `MSDOS.ASM` (then called `DOS.ASM`), plus the FAT12 plumbing and a TRANS86 translator that converted 8080 CP/M assembly to 8086. There is no build system. There is no test suite. There is a `README` Paterson typed in 1980 explaining how to assemble it using SCP's own cross-assembler.
The code's technical interest is real but narrow: nobody is going to fork 86-DOS and ship it. The historical interest is the point. Read end-to-end, 0.1-C is a working operating system written by one person in roughly two months — kernel, shell, file system, and device drivers — and it fits comfortably inside a single browser tab. That's not a romantic exaggeration. The entire `MSDOS.ASM` is under 2,500 lines. The shell (`COMMAND.ASM`) is under 1,000. A modern Electron app's preload script is bigger than the operating system IBM shipped to launch the PC era.
There are two specific things worth reading the code for. First, the FAT implementation. FAT12 — still the file system on every SD card you format for an embedded board — is laid out here in its primordial form, and you can see exactly which design choices were deliberate (cluster chains, the reserved sector) versus which were accidents that calcified into thirty years of compatibility (the 8.3 filename limit, the FAT itself living at a fixed offset). Second, the CP/M emulation. Paterson explicitly designed 86-DOS's INT 21h API to mirror CP/M's BDOS calls so existing CP/M apps could be ported by mechanical translation. The reason DOS's system call interface looks the way it does — function number in AH, parameters in DX — is that it was reverse-engineered to be a drop-in for an 8-bit OS Paterson didn't have the source to.
The community reaction on Hacker News (121 points, single-source coverage so far) split predictably. Retro-computing folks dug straight into the diffs against MS-DOS 1.25 to see what Microsoft changed post-acquisition (answer: surprisingly little in the kernel; mostly the assembler dialect and some IBM-specific BIOS glue). Younger commenters fixated on the licensing — MIT on code that was the subject of a 2007 lawsuit by Digital Research's heirs alleging Paterson copied CP/M, a suit Paterson won. Microsoft's release is not a legal statement, but publishing under MIT in 2026 is, at minimum, the company saying it considers the provenance settled.
The third reaction, the one worth lingering on, is the engineers who pulled up the source next to a modern kernel and noticed they could actually understand it. That's not nostalgia. That's a measurement. The cognitive load of "the whole OS" used to fit in one person's head. It does not anymore, for reasons that are mostly good (preemptive multitasking, memory protection, SMP, networking) and partly bad (twenty layers of abstraction that exist because nobody is willing to delete code).
Nothing, directly. You will not be linking against 86-DOS. But there are two practical reasons to spend an hour with the repo.
If you work on embedded systems, bootloaders, or anything that talks to FAT-formatted media, the 0.1-C FAT code is the cleanest reference implementation you'll find. The Linux kernel's `fs/fat/` is correct and battle-hardened and almost unreadable for pedagogical purposes; Paterson's version is the opposite — pedagogically clear, missing every edge case, and small enough to hold in working memory. It's useful in the way reading the original K&R C book is useful: not because you'll write code that way, but because it shows you what the abstraction is for.
If you work on anything with a system call interface — a runtime, a sandbox, a WASI implementation, an FFI layer — the INT 21h design choices are a case study in how API decisions outlive their context by decades. Paterson chose register-based calling because the 8086 had eight registers and a function call was 9 cycles. That decision is why DOS programs were fast in 1981 and why DOS extenders were a nightmare in 1993 and why DOSBox has to emulate a register-based syscall ABI in 2026. Every syscall interface you design today will look equally arbitrary to someone reading the source in 2071. Plan accordingly.
The MIT license matters too, in a small way. The 2014 MS-DOS 1.25 release used a custom non-commercial license that prevented anyone from actually building on it. The MIT relicense — applied retroactively to the older releases as part of this drop — means hobbyist OS projects, emulators, and FreeDOS can now legally lift code directly. FreeDOS maintainer Jim Hall has already posted that he's reviewing 0.1-C for ideas, though he's cautious about license-laundering modern GPL code through a vintage MIT base.
Microsoft has done this kind of release roughly once every four years, and the company's GitHub posture suggests more historical drops are coming — Word 1.1a (already out), GW-BASIC (out), and there are rumors of Xenix and the original Excel for Mac source being prepared. None of it changes your sprint. All of it is a reminder that the software industry is now old enough to have an archaeology layer, and that the artifacts are surprisingly legible when you crack them open. Read the FAT code this weekend. It'll take an hour, and you'll write better embedded code on Monday.
<a href="https://opensource.microsoft.com/blog/2026/04/28/continuing-the-story-of-early-dos-development/" rel="nofollow">https://opensource.microsoft.
→ read on Hacker NewsI cannot describe to you how jealous I am of the fact that back then writing a few thousand lines of assembly was what it took to launch a successful software company.
Discussion, on the source, at the time (79 points, 24 days ago, 19 comments) https://news.ycombinator.com/item?id=47957494Or on the GitHub clone (162 points, 15 comments) https://news.ycombinator.com/item?id=47946813
wow, they had to OCR it back in from paper printouts> This source code is old enough that it hadn’t been stored digitally. “A dedicated team of historians and preservationists led by Yufeng Gao and Rich Cini,” calling itself the “DOS Disassembly Group,” painstakingly transcribed and scanned in co
I got it up and running in dos box - it still runs smooth as butter: https://github.com/phyous/DOS-1.0
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
It is rare that I say this but, thanks MS! Arguably just as, if not more, important is the BASIC that they wrote. That was what they actually wanted to do. DOS just got them the contract with IBM. For decades MS was really a developer tools company with a side biz of writing operating systems and ot