1966 Lotus Elan FHC: The "Bought-not-Built" car

[2mAn]

Surely you didn't give up on that DIY harness?

Dont call me Shirley, but yes... I don't know what I'm doing so it makes more sense to buy one thats done... Its been frustrating trying to get help (you've been helpful but youre also on the other side of the country) and I would rather spend a little more to get a completed harness VS building one, tuning it and hoping it all actually was done right.The AEM ECU also appears to be a little too 'complex/ advanced' for what I need the ECU to do.

Greenland HOWEVER is making this purchase get more and more expensive... and we're already another week since he said he was almost done... If Im having this dilemma in a few weeks, the whole projects days could be numbered... scary and sad but I will be fuming if I'll have owned this car for a YEAR without ever really driving it.

[2mAn]

While the wait for QED continues, I am trying to continue making progress in the hopes that when I finally get my harness from them, I am only needing to do so. However, the fuel system will likely be the item that waits until I get my ‘kit’ from him to see what I have and what I don’t have. I already bought some items to use related to the fuel system, but they will sit and wait.

Anyways, I ‘smoke tested’ the o2 sensor bung and it didnt leak anything so I felt comfortable installing it with the sensor installed.




Overall, I am very happy with its alignment in the tunnel. Still need to figure out where the hole will be for the wiring to pass through, but thats a problem for another day. With that, I felt it best to take the car down off the rear jack-stands... Hoping to do that to the front soon.



I also was able to get the coolant temp sensor piece to sit flush, but still need to re-tap it so the sensor can properly thread into its hole. At this point, I am blaming AEM for the incorrect info about the thread/ pitch which is advertised as 1/8" NPT...







YouTube video if you'd like to hear my nonsense...

[2mAn]

Week is almost over and still no update from QED…

In a “fun” turn of events, my work had me do some AI training which forced me to interact with Microsoft Copilot, and once that training was over they suggested practicing my prompts so I figured that I could see if it could draw me a wiring diagram for the AEM ECU. I havent bothered continuing the AEM route once I committed to the QED setup, but while I wait… why not.

Heres as far as I got, along with my notes of “i dunno what to do here…” which was a modified version of the provided AEM model


I had Copilot ‘read’ the entire 406 page PDF manual, gave it the stock image I worked off of, and this was the best it could come up with. I might start again from scratch to see what it can do

[econti]

Disclaimer: not an expert.
That AI diagram is fucking atrocious, do not trust any of it. Do not use AI at all for generating or modifying a wiring diagram, it will not work.
Your diagram looks like you're pretty close. Yes, the injectors just need a shared positive power, the ECU switches the ground to make them fire. Are you using coilpacks? I assume so?

[Northern]

Nah I say go for it. For science.

I want to see how that 4 wire cam sensor setup goes, which spelling of ignicton/ignittion comes out on top, and whatever the cursed shit in the bottom right corner is.

[roguetoaster]

That's impressive, truly an excellent use of resources to create that helpful diagram.

[McGyver]

Damn, the AI diagram does look pretty clear. Maybe you should follow it?

My company just had AI training this week too. We're also supposed to use copilot, but apparently our IT department isn't happy with the current state of the GCC-High, so they won't roll it out. So I guess we're supposed to tinker with it on our personal accounts?

One thing I took from the training is that AI needs to be given a framework for solving a complex problem. This was really surprising to me and was somewhat disappointing. It may also help to turn down the "temperature" (creativity) since your use case needs to be applied to problem solving vs content generation. I would have the AI define the systems that your car needs, then map out each system, then give you a consolidated wiring diagram. It would be best if this was accomplished by running some sort of code vs chat to force repeatability. You may also have a different AI review the output and generate suggestions, then put those suggestions into copilot.

I had put together a spreadsheet to calculate finished goods stock levels so that we could ship 90% of orders same-day for those part numbers. My calculation gave me something like needing 81 in stock. I ran the spreadsheet through copilot and it told me my numbers were way off, that i needed 200-300 items in stock for 90%. When I asked it to show how it calculated that, it's numbers were all wrong. I corrected it, it agreed, and then came to my calculation. I think it might have worked if i did a better job of getting it the raw data and building out the prompt. It was fun to play with, but ultimately a waste of time.

[roguetoaster]

It was fun to play with, but ultimately a waste of time.

Aye.

[2mAn]

It was fun to play with, but ultimately a waste of time.

Exactly what this was... if I have more time to burn today I might try it again starting from scratch.

The initial prompt was "design a wiring diagram using the aem infinity ecu to use on a lotus twin cam motor" several prompts later, it produced what I showed above.

I think that sadly as much as I dont want to use AI, I need to develop my skills 'writing prompts' so that I can use this tool. I (and likely most of us here) have ~25 years or more ahead of us and as much as Im a late adopter and denier to new tech I need to jump on this so that I can feed my family, and keep buying stupid cars

[Northern]

AI is brutal. Tempting to use it as a shortcut to using modern enshittified google, and I've done it before with extremely limited success.
I find it gives you some summary of what's generally in the top few hits on google, but struggles to go deeper.

If you ask a specific question that wasn't obvious from the previous response, it struggles to provide anything close to an adequate answer.

It may be ok at clarifying things that have a lot of conflicting information on the internet (specific example was asking about GM platforms with aluminum 5.3s and 6L80e, but I'm very skeptical of what it said)

My litmus test is to ask it about something I know about. The inaccuracies stand out like a sore thumb.
If you can't figure it out in a minute of googling, it can't either.

[uturn]

My take on AI... a hundred idiots in the room claiming their wrong answer is correct will drown out the one guy with the correct answer. It's just crowd sourcing information... so I take it with a grain of salt.

[McGyver]

Exactly what this was... if I have more time to burn today I might try it again starting from scratch.

The initial prompt was "design a wiring diagram using the aem infinity ecu to use on a lotus twin cam motor" several prompts later, it produced what I showed above.

I think that sadly as much as I dont want to use AI, I need to develop my skills 'writing prompts' so that I can use this tool. I (and likely most of us here) have ~25 years or more ahead of us and as much as Im a late adopter and denier to new tech I need to jump on this so that I can feed my family, and keep buying stupid cars

I really wanted to resist using AI, and did for a while, but ultimately came to the same conclusion as you: I can refuse to use it, but someone else is learning how to apply it and all industries are pushing it for productivity savings. I can either get on board now, or get left behind.

I use it often when I need to create content. I regularly need to write work instructions or develop PowerPoints to lead improvement events. I find it useful to take my brain-dump concepts and turn it into something organized and digestible. I find that it saves a lot of time that I would otherwise spend on editing as I go along. I've also used it a ton for job applications. I'll find something I like, get it to summarize the post, then compare my resume to it and point out the gaps. I provide examples to fill the gaps, and it provides text for the resume. It's also great for responding to recruiters when I don't want to put the effort into something that probably won't pan out.

For your use case, I would focus on adding some extra context. Specifics about the engine, sensors, etc. Adding the full manual is really great for context, but remember the "context window" of what AI will remember in the chat. If the manual is too long, it'll overwrite that "memory" and you won't know. For Copilot, they suggested putting all the context into a notebook, then reference the notebook in your chat. Hell, if you can download this thread, the whole thing could be in the notebook. Then you would get it to identify the major systems, the components in the system, the wiring diagram, and eventually you could probably get it to give you wiring lengths based on a schematic of the car.

[2mAn]

Well… an order has been placed. I began communicating with DTAfast on Monday, after another missed “Its going out today” (Thanks for the suggestion, Tim!) They have an understanding of what I’m after and I should have the harness & ECU very soon, once we work out some of the final details.

March 22, 2026 is the date of the Queens English car show, and is the event I am shooting to have the car ‘debuted’ at. Lets see what I can do…

T-minus 44 days. Yikes

[MrBurgundy]

Any wagers?

Will it run in 2026?

Will it make it to the English car show? Will any English car make it?

[roguetoaster]

Everybody comment on his first startup video with "You WOT mate?"