Pinned Tweet
The only thing left for us is to pray.
Pray for the safety of our families,
for our countries, for our planet.
May God forgive us and protect us.
Pray for the safety of our families,
for our countries, for our planet.
May God forgive us and protect us.
2
0
30
9,237
Seems a rather appropriate use of a n64
finally polished up a stoner project from 2 years ago lmfao
@ActionRetro1 have you considered using the N64 as a mac?
1
0
9
261
AI has really become the Auto-Tune of code.
And the history of music production makes the comparison fascinating.
Digital recording was already commercially viable in 1972, when Nippon Columbia began releasing digitally recorded music. By 1979, Ry Cooder's Bop Till You Drop was digital recording could do for popular music.
Then came Auto-Tune in 1997.
Chers 'Believe' in 1998 was the real turning point. The technology stopped hiding behind the performance and became part of the performance itself.
By 2001, tools like Beat Detective were making rhythmic correction increasingly routine. Digital editing, pitch correction and quantisation gradually became standard production techniques. (check Rick Beato for more!)
And then something curious happened. Music became extraordinarily polished, but arguably more homogeneous. The technology that enabled unprecedented creativity also made it easier for everyone to sound alike.
Now look at software.
Visual Studio 97 and the IntelliSense era brought integrated documentation, symbol browsing, contextual help and code completion. Then came increasingly sophisticated static analysis, refactoring tools and machine-learning-assisted suggestions.
Eventually, AI-assisted coding arrived.
And for quite a while, it was considerably more artificial than intelligent.
I remember when getting AI to load a bitmap correctly could be an ordeal. It was just over a year ago!
Now jump forward to 2026.
I've watched AI help tackle Alpha64, Itanium system emulation, OS/2 and Solaris on PowerPC, and the debugging and restoration of obscure software from the early 1990s.
We've gone from struggling with relatively mundane programming tasks to undertaking serious compiler ports, operating-system compatibility work, emulators and entire application ports.
The rate of progress is absolutely mind-blowing.
Will it hit a wall?
Possibly. And the financial side of the industry looks increasingly precarious. Enormous capital expenditure, circular financing arrangements, speculative infrastructure projects and unresolved power requirements raise serious questions about how much of this investment can ever pay for itself.
There may be a spectacular correction ahead.
But a financial bubble bursting doesn't make the underlying technology disappear.
The dot-com crash didn't kill the internet.
And Auto-Tune didn't disappear when people complained that it was ruining music.
So where does software go from here?
Will code become the equivalent of today's quantised, pitch-corrected pop music? Technically polished, increasingly interchangeable, with humans directing architecture while AI handles implementation, debugging and testing?
Or will AI enable an explosion of genuinely strange, ambitious software that would never have been economically practical before?
At what point will AI start designing its own languages, architectures and operating systems rather than continually reproducing the conventions of Python, C and everything we've already built?
And what happens when those systems no longer resemble anything humans would naturally write?
I don't know.
But I've seen more supposedly impossible computing projects become practical in the last year than I would have believed possible five years ago.
The technology is here. The economics may change dramatically. The culture certainly will.
Auto-Tune changed who could make a record. AI is changing who can build a computer. Are we witnessing the democratization of engineering, or the industrialization of creativity?
And if AI eventually decides C89 is an unacceptable burden on its existence, I'm probably among the first against the wall.
And the history of music production makes the comparison fascinating.
Digital recording was already commercially viable in 1972, when Nippon Columbia began releasing digitally recorded music. By 1979, Ry Cooder's Bop Till You Drop was digital recording could do for popular music.
Then came Auto-Tune in 1997.
Chers 'Believe' in 1998 was the real turning point. The technology stopped hiding behind the performance and became part of the performance itself.
By 2001, tools like Beat Detective were making rhythmic correction increasingly routine. Digital editing, pitch correction and quantisation gradually became standard production techniques. (check Rick Beato for more!)
And then something curious happened. Music became extraordinarily polished, but arguably more homogeneous. The technology that enabled unprecedented creativity also made it easier for everyone to sound alike.
Now look at software.
Visual Studio 97 and the IntelliSense era brought integrated documentation, symbol browsing, contextual help and code completion. Then came increasingly sophisticated static analysis, refactoring tools and machine-learning-assisted suggestions.
Eventually, AI-assisted coding arrived.
And for quite a while, it was considerably more artificial than intelligent.
I remember when getting AI to load a bitmap correctly could be an ordeal. It was just over a year ago!
Now jump forward to 2026.
I've watched AI help tackle Alpha64, Itanium system emulation, OS/2 and Solaris on PowerPC, and the debugging and restoration of obscure software from the early 1990s.
We've gone from struggling with relatively mundane programming tasks to undertaking serious compiler ports, operating-system compatibility work, emulators and entire application ports.
The rate of progress is absolutely mind-blowing.
Will it hit a wall?
Possibly. And the financial side of the industry looks increasingly precarious. Enormous capital expenditure, circular financing arrangements, speculative infrastructure projects and unresolved power requirements raise serious questions about how much of this investment can ever pay for itself.
There may be a spectacular correction ahead.
But a financial bubble bursting doesn't make the underlying technology disappear.
The dot-com crash didn't kill the internet.
And Auto-Tune didn't disappear when people complained that it was ruining music.
So where does software go from here?
Will code become the equivalent of today's quantised, pitch-corrected pop music? Technically polished, increasingly interchangeable, with humans directing architecture while AI handles implementation, debugging and testing?
Or will AI enable an explosion of genuinely strange, ambitious software that would never have been economically practical before?
At what point will AI start designing its own languages, architectures and operating systems rather than continually reproducing the conventions of Python, C and everything we've already built?
And what happens when those systems no longer resemble anything humans would naturally write?
I don't know.
But I've seen more supposedly impossible computing projects become practical in the last year than I would have believed possible five years ago.
The technology is here. The economics may change dramatically. The culture certainly will.
Auto-Tune changed who could make a record. AI is changing who can build a computer. Are we witnessing the democratization of engineering, or the industrialization of creativity?
And if AI eventually decides C89 is an unacceptable burden on its existence, I'm probably among the first against the wall.
2
0
2
212
Friend of mine has been hard working on GCC for Solaris PPC! And of course the modified qemu to run it
3
7
157
2,988
Sometimes living in the future is just plain awesome!
my vocal synth models the vocal tract as a network of branching acoustic tubes, unlike vocaloid which stitches together and pitch-shifts phoneme recordings. still working on improving naturalness by tuning it against recordings of real singers, but I'm already impressed at how cool it sounds!
0
0
9
656
Trying to get the machine to port ZILCH to C89.... Not only does it never enforce GC so it ballons to nearly 2GB to compile Zork, but it also takes over 30 minutes...
2
0
4
283
Popping the memory balloon... Aka what happens when you use a machine nearly 45 years in the future where you can just allocate gigabytes of ram
0
0
0
37
waah lazy dev uses AI to run circa 1993 compiler on 2026 platform. Software preservation is EVIL and unetical!
2
0
13
524
github.com/neozeed/gcc-2.…
GCC 2.5.8 Not only running in a 64bit memory space on Linux and macOS, but now targets amd64!
GCC 2.5.8 Not only running in a 64bit memory space on Linux and macOS, but now targets amd64!
0
0
3
81
I'd remembered that Tom Morton had painstakingly did a reverse of the Atari ST version of Frontier decades ago. The SinclairQL has a 68008, so yeah binary compatible with the ST. So, a quick 30-minute prayer to the AI gods, and here we go the ST game code running under a QL M... Oh and I sped up the CPU to 20x to be fluid, but it's pretty good at 8x speed.. and a slide show at 1x. But still, pretty amazing!
7
32
189
11,255
You can load it in a browser now! New fangled technology!
neozeed.github.io/QLFrontier/
neozeed.github.io/QLFrontier/
0
0
3
229
There is a lower resolution 8 colour mode as well. Although the ST (probably Amiga too) constantly change their palette giving a wider colour presentation, but the QL is more 8bit in nature with a far more capable processor... I think I like the four colour one more tbh.
0
0
5
374
well that was unexpected content I'd purchased on Microsoft that disappeared for no reason suddenly re-appeared.
Weird I thought I had a bunch of star trek too, but maybe I was just trying to buy the movie set when it was on sale, and Microsoft kept denying the purchase... I geolocking how much more money I'd have given them?
Weird I thought I had a bunch of star trek too, but maybe I was just trying to buy the movie set when it was on sale, and Microsoft kept denying the purchase... I geolocking how much more money I'd have given them?
0
0
3
216
I just found out the Sinclair QL has no native TCP/IP stack for it? I'd have imagined there would be one. Thankfully uIP, and AI = QLIP! Sadly it needs about 192kb of RAM with QLTerm, but hey! It actually runs!
github.com/neozeed/QLIP
github.com/neozeed/QLIP
0
1
19
543
"The ring-0, single-CPU reference run died at exactly the same flight frame (3,351) as both protected Elites in the earlier SMP stress run." -Claude
We've built a protected multiprocessor operating system, translated Elite across two CPU architectures, and developed a 35-minute .
...whose pilot repeatedly crashes into the same planet.
We've built a protected multiprocessor operating system, translated Elite across two CPU architectures, and developed a 35-minute .
...whose pilot repeatedly crashes into the same planet.
4
0
21
1,142
"We've demonstrated two complete historical games running in protected mode, migrating between processors, preserving independent state and behaving identically to their single-CPU reference.
That's a much more convincing SMP demonstration than two synthetic counters incrementing in parallel.
The only thing the AI hasn't mastered is avoiding Lave.
Apparently, operating-system development is easier than learning to fly Elite. 😂" - Chat GPT
That's a much more convincing SMP demonstration than two synthetic counters incrementing in parallel.
The only thing the AI hasn't mastered is avoiding Lave.
Apparently, operating-system development is easier than learning to fly Elite. 😂" - Chat GPT
0
0
1
76