It "technically" can - i used Win95 on an AMD 386DX with 4MB of RAM back in the 90s for a little. Most of the time was spent waiting for the computer to do something while the HDD was constantly working.
The Pentium MMX with 32MB i replaced it with was MUCH faster, practically everything was instant.
I tried installing it on my 20mhz 486 IBM but the install never would finish even though the minimum specs listed 386. I always thought that was being too generous. I did use it on faster 486's at some point in school (I think 66mhz or faster?)
Back in the mid 90’s I got it to “run” on a 386, but it was miserable. It was basically wondering if it was hung or just really slow on every command or mouse click.
Yes. I remember one of the MS guys saying the goal for win 95 was win32s equivalent performance on a 386 with 4 Mb ram. But win32s performance on that hardware was crap :)
supposedly with a minimum of 4M RAM even! [1] I have a machine with twice that clock and twice that RAM that a friends parents wanted to toss out[2]. I spent the better part of a day getting Windows 3.11 and Office 4.3 to run. Playing around with it for a bit, I highly doubt that Windows 95 on minimum requirements would be anything but endlessly frustrating. I was not masochistic enough to try.
I ran Windows 95 on a 386SX-16 with an 80387, with 5MB of ram for a year as my main machine. It ran about as well as 3.1 did. The only real drawback was the disk space required for 95 over 3.1 + DOS.
Good question. I ran Slackware Linux and DOS 5.0 on mine. I think Win95 was my Pentium 100 MHz already so it sounds a bit sus, but the FAQ section of OP had items about Win95, that is where I got the idea from.
We had Win95 running on a 25MHz 4MB FPRAM 286 back in high school. Poor teacher, half of first period was roughly how long it took to boot up, only for them to exit into DOS mode because the grading software and report card software was DOS-native and would hang if you tried running it under the GUI.
> Windows 95 had a beta build for 286 systems, which is what this system 'ran.'
I think you are confused. There was an Windows 95 build 286, but despite its name it didn't run on a 286 – it was just an incrementing counter which happened to have reached that number.
At some point early in Windows 95's development, Microsoft dropped the standard-mode only 286 kernel, KRNL286.EXE. They also dropped 16-bit-only standard mode from KRNL386.EXE; but KRNL386.EXE required a 386 even for standard mode.
The earliest leaked Windows 95 build, build 22, still has KRNL286.EXE in it, but it is unchanged from Windows for Workgroups 3.11, and doesn't actually work – other Windows 95 components require new APIs which Microsoft added to KRNL386 but didn't bother adding to KRNL286. By the time we get to build 286, KRNL286 was long gone.
In principle, you could likely patch early Windows 95 builds to run on a 286, but I don't think anyone has actually tried. At a minimum, you'd need to add the missing APIs to KRNL286.EXE
Taking you at your word, I suspect what you may have actually saw was https://en.wikipedia.org/wiki/Calmira – an open source project which gives Windows 3.x a Windows 95-style interface – or if not Calmira, other people implemented the same idea independently, e.g. the shareware Mask98 package – http://www.win3x.org/win3board/viewtopic.php?t=1262&view=min – however, this had nothing to do with Microsoft; Microsoft could have done something like this, but they had no commercial interest in doing so.
> Bet you don't even know about Windows 97.
That is just an unofficial/unlicensed third-party distro of Windows 3.11 with Calmira bundled.
Wow, I actually had this exact idea. I was specifically curious as to how well a given LLM could understand a DSL that hasn't changed much in a couple decades and doesn't have nearly as many examples to learn from online. Seems like it did alright, all things considered.
Not to sound like a contrarian, but no. They would have thrown it in a faraday bag, cloned the device, and then would use one of the many existing forensic tools available to crack the pin.
That's crazy. Here in the UK border police can question you for any reason (no need for suspicion) and if you refuse to answer questions or give them access to your devices you'll be charged under terror legalisation.
> Under Schedule 7, a police officer with the mandated accreditation does not need prior authority or suspicion to stop, question, search, or if necessary, detain someone. However, they may only stop and question a person for the purpose of allowing a determination of whether that person appears to be someone who is or who has been concerned in the commission, preparation or instigation of acts of terrorism.
How this is abused on the other hand, I presume, might depend on your skin shade.
The charges are the punishment. You are imprisoned, risk losing your career, risk losing reputation, and still have to hire a lawyer. The costs are incredibly asymmetric. This leads to the majority of people complying, which only emboldens the state.
I intuitively and perhaps naively thought that Claude using the LSP server would negate a lot of grep use. Is this tool solving the same problem or something else?
LSP is for when you already know the symbol and want an exact answer, like go-to-definition or find-references. graft is for the step before that: you ask "where does auth happen" and it hands the agent the right files and lines to start from, so it reads a lot less to get oriented.
so graft solves for making claude code understand your codebase every session.
And I feel you can save a lot more tokens and also increase accuracy by just switching to new claude session without worrying about your codebase context. (doesn't solve for the session context yet, but we are still figuring out how can we solve for that. Maybe hooks that can manage your CLAUDE.md)
For a while everyone was saying sub agents is how you save tokens, use lower quality models with limited context to do simple parts of the job after a smart planning agent has put it all in place. Is that no longer true or is this just the result of sub agent being used at the wrong time?
No, you can definitely configure low cost search and apply subagents. CC and Codex do not. Not sure if this is to improve the reliability of their subagents, or just a play to increase user consumption.
> Any time I notice something in code review and have to get the agent to fix it.. I throw it on the list! My list is like 200 items now.
This is a gripe I've had with AI tools for a while now. Though it's gotten somewhat better in time, but we don't really know what to expect from the tool in terms of quality. Ex. I'd expect a human engineer to probably not use a brand new assertion library for a new test when there are 200 tests using an existing one. But Claude has done this to me multiple times. So I have to add yet another item to the list, like you have, and tell it to look for testing conventions before writing. But, there is plenty we don't have to tell it, like what a function is or a test should probably cover the change in the diff. But we don't really have a list of what things are on each side so we're just left to sort of hunt and peck to build a viable solution.
You're forced to document every check and thought related to development until you're wiring a manual on how to be laid off (unless self employed or otherwise critical).
Which is fine. One should write quality docs and all that. But it seems exactly like training my offshore contract replacement instead of building tools for my own use. Except that income doesn't even go to a human just a black box datacenter company and shareholders.
I find project-wide or subpackage-wide (for sufficiently large projects) CLAUDE.md's that document patterns (including for tests) solves this.
/init will make a project-wide one, or you can instruct it to "Create CLAUDE.md in any sub-directory that is sufficiently complex" then modify from there.
reply