The biggest mistake computer companies made in the 70s an 80s was to ignore software and backward compatibility. Tandy killed themselves by making Model 1 and Model 2 incompatible. How hard would it have been to make the VIC-20 run Pet software? Apple Macintosh would have been more successful early on if they had included an Apple ][e on a chip and allowed it to boot into Apple ][ mode like Commodore did with the C64/C128. The TED machines flopped because they had no software and were no more capable than the C64. Atari ST and Amiga also struggled because they began with very little software.
The second mistake was to delay networking. Computers are far more valuable when connected. Commodore was closest to getting it right when they connected a bunch of Pets to a pair of disk drives. Acorn's tube architecture for connecting multiple processors was brilliant. It should have been ubiquitous. It would also have helped with the new-machine-no-software problem.
There is no reason Amiga's indirect color palette could not have existed in every video display interface built in 1979, or even earlier. Three ubiquitous 7489 chips and some glue will give you 16 of 4096 colors with a simple circuit. There is no excuse for CGA's grotesque colors.
Higher resolution video was hampered by the cost of memory and because we didn't know how to deal with EMI at the frequencies we needed. Memory was starting to get cheap enough by 1980 to allow 640x480, but higher resolutions needed purpose built-monitors and high-bandwidth cables. Only purchasers of high-end CAD workstations could afford the extra expense.
Atari ST made a hit by including MIDI ports in the base machine: Another win for ubiquitous connectivity.
Specify high-quality capacitors in every design. Every battery that could explode or leak is isolated on its own board and placed where it cannot destroy critical circuits.
homebrew 6502 with 4-16k RAM (code name: uqim)
split ortholinear keyboard (like keyboard.io Atreus)
video interface like Processor Tech's VDM-1
Kansas City cassette interface
receipt printer for hard copy: write fixed-point budget software to manage family finances.
Turn the main board of the previous system into a controller for a SA-400 mini-floppy. (code name: Kepler)
Build a new main board with more RAM. (code name: Archimedes)
Connect the two with RS-485 interface, (call it CFLL: common fast local link. code name: Hermes - god of travelers, messenger of the gods.) similar to Commodore's IEC serial port, with a packet-oriented protocol that generalizes to more than just disk drives. Makes SCSI and ATA unnecessary. Vastly simplifies OS code because everyone writes to the same API and most of the work is done by the interface card.
Build a Z-80 card that will run CP/M (code name: Zeus). Create my own CP/M replacement -- called Ubiqs -- that will also run on 6502
Buy an Apple I and adapt Zeus to it. Create a CFLL card.
create a plotter to help me design and produce PCBs.
Improve the cassette interface to include motor control. Write code to copy data between disk and cassette. (Sell this to software houses for duplication, and/or offer software publishing as a service myself.)
Write VisiCalc for Apple ][, CP/M, and Ubiqs (Yes, I know: VisiCalc wasn't released until May 1979. As long as I am dreaming, I may as well dream of writing a killer app.)
Use the profit from VisiCalc sales to fund hardware development.
adapt Zeus to Apple ][
create CFLL cards for Apple ][, S-100, and TRS-80.
I'm not sure whether it would be easier to port CP/M VisiCalc to TRS-80 or to adapt Zeus to run inside. I might do either or both.
maybe adapt Archimedes to run inside a TRS-80 as an attached processor. (Will run VisiCalc/Ubiqs)
Sell standalone disk drives with Kepler board and 4 CFLL ports as a universal NAS. (code name: Euclid) (think: Commodore disk drive that will attach to any computer via CFLL.) Apple ][ users need not wait for Disk ][ (mid 1978). Schools and businesses can easily share files between TRS-80, CP/M, PET, and Apple computers.
adapt Zeus to Commodore PET. adapt CFLL card to PET.
add 2.4 GHz wireless capability to CFLL cards - create WIFI standard with encryption and zero-knowledge authentication built into the base protocol.
develop video display card with capabilities similar to Commodore 64 (code name: Astro)
develop a sound/MIDI board (code name: Rosie). Dedicated audio RAM much like video RAM on VERA. 8-bit stereo DAC and ADC. 4-voice synth with ADSR modulation.
prototype 8086 compute engine to attach to my main board the same way Zeus does. (code name: Hades - programming the 8086 seems like living in the underworld, but 64k is getting cramped for the applications I want to build, so we deal with it until the 68000 is ready.)
port VisiCalc to Hades.
sell Hades as an accelerator for VisiCalc. UI and I/O runs on the host computer, compute happens in Hades as in Acorn "tube".
finally release Archimedes workstation to the public with Astro, Rosie, Zeus, Hades
Update NAS to support hard disks and RAID. This will be a boon to software companies everywhere.
develop upgraded video display card with VGA-level capabilities (code name: Elroy). Video input for frame grabber and genlock.
develop a digital RGB monitor to go with Elroy (code name: Iris)
develop upgraded sound/MIDI board (code name: Judy). 16-bit DAC and ADC.
release 8086-based workstation (code name: Ptolemy) with Judy. Includes 6502 processor to run all legacy software. Runs CP/M software under Ubiqs. 80-column Astro video display for fast text. Elroy and Iris add-on for those who can afford it.
(also at November Comdex) release Ptolemy all-in-one luggable like the Osborne 1 with a 9-inch CRT. Designed to fit size requirements for carry-on luggage. (code name: Rubius -- reminds me of Hagrid. big and awkward, but also strong and friendly.)
create CFLL card for Apple /// (if the Apple ][ version doesn't work)
68000 compute engine (code name: Poseidon - as the waters cover the earth...) cf. DTACK Grounded.
bull case: IBM licenses Ubiqs instead of MS-DOS (which came from Seattle Computer Products. Oh. Maybe I should *buy* SCP.)
CFLL card for VIC-20, IBM PC
adapt Zeus, Hades, and Poseidon to attach to IBM PC (Why Hades? To make IBM look silly. It will already be more powerful than the PC. But it works: the host computer is just a terminal, and the software has already had 2 years to mature running in older 8-bit hosts.)
68000 workstation (code name: Copernicus)
Ptolemy-based laptop with LCD screen (code name: Albus)
CFLL card for C-64 (would the VIC-20 version work unmodified?)
adapt Archimedes, Zeus, Hades, and Poseidon to attach to C-64
2-D raster graphics accelerator (blitter) (code name: Demeter - goddess of agriculture. Think: rows of grain)
Create laser printer and launch desktop publishing industry.
create CFLL card for Apple Lisa. It won't matter, but I should do it anyway.
create CFLL card for Apple Macintosh (Experience I have gained with hard disks and RAID will make NAS users very happy and Apple hard drive sellers very sad.)
compute engine based on ARM or ARM-like core (code name: Galileo) The new low-power RISC chips will be important for the next big leap. Motorola and Intel will be sad. The new engine will plug in seamlessly to the architecture I developed over the previous eight years.