Oct 13

 Sluggish today.

Sleep has been cloying, muddied, deep. Each approach of the surface of wakefulness has come with an ache in my torso, that sucking, straining, ill feeling. It is an odd feeling. It feels "ill". Feverish without the temperature. A sickness. Without some of the paraphenalia, no stuffy nose, no sore throat, no temperature. But feeling ill nevertheless.

And with that sticky, suffocating sleep. Waking has been mired in treacle and fog. Today I have struggled to do much useful at work. A couple of queries. A couple of checks. But the engine has failed to start. Despite me gritting my teeth. Repeatedly kicking it. It wouldn't start. And eventually. I fell into the inevitable. And slept. And slept. More treacle. More fog. Deep in some exhausted asleep state. When I circle in the shallow waters of still asleep, but somewhat aware I can feel it. An enormous weight and that illness. It takes a great deal of effort to surface out of it. It is easy. No effort. To give in and sink again. I wonder sometimes if that's one of the ways I might go. Sinking into those depths. A couple of years ago now with the whole opiate based painkillers and anti allergy meds. That hole was very deep indeed. Very hard to crawl out of. And I wondered then if that was a little push down that hill.

Anyway.

Today has been treacle.

The last month the AI apocalypse has finally, slowly, been getting its feet underneath it. We are moving from the fucking about phase. Into the finding out phase I think. The speed of the flip is incredible. The last month has highlighted issues. The last two weeks accelerated those issues, the last week set those issues on fire. And the last few days....

None of this is particularly revolutionary, at least, in terms of AI development. None of it isn't something we couldn't see coming. It just seems that of late. The pieces have been put together. Implementation inevitably lags innovation. The innovation has been sitting there for sometime. Getting better. By leaps and bounds. And now, the implementations are catching up. Or perhaps more pertinently. The implementations themselves are starting to be written by the AIs themselves. At their breakneck speed. And the dam has started to burst.

Where it ends is anyones guess. Educated guess. Things are going to change. Economies are going to shift. Realities are going to hit. 

Last week Adobe got a shot across its bows. A badly aimed one. Not a well crafted one. But an incredible shot nevertheless. Someone reverse engineered the adobe suite of software. And made it open source. In a week.

A warning.

None of this model works anymore.

A new dark meme has popped up delighting in the gallows humour. Depicting a robot standing over a person and telling them they have been chosen to be open sourced. And not to resist.

I spoke to Andy today about it. He is behind the curve on this. But open minded. And does have an idea of where it might end up. He's just... lagging... because the speed of this is unprecedented.

Last week for the first time we got a request from a client. Non IT. Non technical. They had written themselves an app they said. With chatGPT. Can you help install it on our server ?

No. Was our answer. We are not putting some unverified, unvalidated security risk onto proper hosting.

And this. Is going to be the first of many.

Andy is of the opinion no one knows what they're doing. Who will monitor security. Out of date libraries etc.

This is thin in my opinion. Very thin. *We* don't do that for half out human authored stuff. Some of the work I've done in the last few months has been exactly this - dragging servers that should have been replaced years ago into the light and back into supportability.

And. This too. Is just a tiny easy task for an AI to succeed at. Step 5. Check security implications. Run pen test.

No big deal.

Coding as we know it. Development as we know it, is dead. The industry on that side is done. No doubt, for a while at least, there is going to need to be institutional knowledge. Accountability. Checks. Validation. Understanding of bigger pictures. But all of it in my opinion is thin. And capable of being automated out. Within reach of an AI.

Elsewise.

I have given a tiny bit of thought to therapy on Monday. I have a small amount of sympathy with the therapist. Just wanting to work on what works. Avoid conflict. Out of their depth. I get it. But. On the other hand. You have to be able to admit your mistakes. If you can't. You will always be stuck at whatever level that is. Incapable of even acknowledging a problem. Let alone moving on. And for a therapist, the irony is real. This should be their bread and butter. They should know all about psychological sticking points. Physician heal thyself. Ultimately. If they can't admit it. Won't admit it. That's ok. It gives me a valuable piece of information. That we have hit your competence ceiling here. And you can go no further. And sure. In theory. Perhaps there are other niche places you could find where that wouldn't be an issue. Literally. Just a sympathy reflection. Without analysis. Without kicking the tyres and making sure I am not losing the plot. But. That has then become so thin of a therapeutic service. No insights. No checks. Incapability to work together without ego getting in the way. What's the point ? 

I think ironically, in the end. The therapist ends up disqualifying themselves from further work because of an admittance of incapability. Not explicitly. Not honestly. But because of the wall of ego they cannot climb over. And when it is shown in stark terms, there is the wall. Say it. Or do not. And they shy at the hurdle. Ok. Information received. You cannot complete the assault course. Thank you for your time.

Beyond that.

Not sure what I am doing here. Existing. And my need to do things with computers diminishes. As much as I like working with AIs. I can see. That they make so many endeavours... pointless. I looked at some code today that I had written. Why bother I thought. The AI could do this. In a fraction of the time. Probably to a better standard - at least in terms of documentation and checks etc, if not actually better code execution wise.

I need to run a full project test against an AI. I am pretty sure it can complete it. But I want to see how it does. Plot out exactly the steps and implications from a professional consultancy point of view. A proper project guide and practices thing. Rather than a loosey goosey internet demo of something 80% there. And then. Consider. What it means for the future based on the results.

I have an odd feeling. For sometime. That I am living through the end times. An end time. I can see the parallels with the industrial revolution. I can feel and see what the fear and uncertainty and anger must have been like back then. It's present now. The world is changing and upsetting a great many paradigms. This time. It's not guaranteed that anything replaces it. It could be the hollowing out of many things. Not simply a change. But an emptying. In a utopia. A cause for celebration. In the capitalist dystopia. A cataclysm.

Eh well.

One day at a time.

With everything. Health. Age. AIs. Climate change.

Get a deckchair. Enjoy the supernova. Be zen about the beginning middle and end of all things. Change or not. Nothing lasts forever.