5 Days No AI

Why this challenge ?

I cannot say I am the heaviest user of AI since I lack the stats to back that up. But I am sure anyone that uses it can tell if they are usage is getting out of hand. And for me, that sentiment is always strongest when I feel my productivity levels are dipping too much. When I feel simple tasks are becoming unusually harder to do. So I was like, how about a simple challenge this week. 5 Days No AI.

But isn’t it too extreme ?

I mean, I have tried 1 month challenges and those didn’t go so well. But this is just just Monday, Tuesday, Wednesday then.. Friday. How hard can that be? Well. I made that decision yesterday and I have already used Grok quite a number of times this morning. But It’s still Monday so I guess I still have a chance. How am I going to do it? I mean, I won’t be sharing my search history but I guess I’ll share my strategy just in case someone , you for instance, ever wants to do the same.

How will I get what AI was giving?

1. Validation

Sharing every thing to get an opinion.. Nope, praise. “Oh, that’s so well done”. Now it seems to me wanting validation isn’t a problem. That’s a human trait. The problem is addiction to a constant stream of an unhealthy version of it. From who? From some psycophanting AI.. Nope. That should go. The people I talk to will give me enough of that.

2. Assistance with thinking through hard problems

After you have been using AI heavily for a while (as a crutch), your definition for hard gradually loosens. At least I think mine has been. The effect? More need for a ’thinking’ partner. Till when its doing all the thinking while you sit their watching as its dumb partner. Sitting through hard challenges (coding block, error messages) solo is going to be a hard one. But I think it is a valuable skill. I had a similar experience challenging myself to debug C++ compilation errors with no stack overflow but just the CPP ref before. It was a pain. But it was interesting how weeks later it was just how I did things. And I easily found my way around problems after that. I miss that. So there won’t be much outsourcing for this role. I’ll resort to docs and community forums. I guess that’s a step above AI reliance.

3. Getting quick tutorials on libraries whose docs I don’t want to read

Sometimes I need to use a library but I’m not ready to read all about it all hunt for the specific part that addresses my need. But that’s not the only reason I’ve been avoiding the docs. After I realised AI could give me exactly what I needed or at times a good enough summary to get the work done, I was like.. Documentation reading – into the trash!. There’s a skill in being able to find what you need from the docs. And I think if one does not use them often, they never get to the point where they can do what the AI does.. quickly find a relevant section for their need. I do this very often with the man pages. /[whatineed] and boom! I got it and am out. So how about I invest time in becoming as good with the docs. I have a growing library of offline docs for the tools and libraries I use most often and I guess it’s time to put those to some good use.

4. Random prompts for nothing

You ever opened your phone not because you needed something but .. just because it was there somehow wanting to be opened. Something addictive about these LLM’s I can’t put my finger on. Some times I’m clicking the ‘Claude’ or ‘Grok’ icon, or typing that url into the search bar just because that’s how easy it is to get them cigarettes. I decide on what I want to do with it when the prompt appears. What is that? I don’t want to know. I know one thing for sure. It ain’t healthy. So there won’t be outsourcing for this. This time can be given to writing code or some other healthy endeavour.

Success metrics **

How much AI usage to pass the challenge?

So like all well written goals, this goal needs some well defined criteria for success. So I can know square and clear what the outcomes will be at the end: A win or a loss. I need to have some defintion of a win. All my AI usage is with remotely hosted LLMs and most of my internet consumption is on my PC. Which is a plus. I have an extension that exports search history. The exported data is in Json format. It includes every url visited, the timestamp for when the visit happened and the number of visits. There’s quite a lot you can do with it.

But I’m interested in analytics for how often I am using LLMs.

The ones I use are:

• Grok
• Claude
• Deepseek
• ChatGpt

So I will export the data on d-day and this week will have a challenge for writing a python script that given a hostname, can tell from exported data how many times I visited it. Success will not be binary but in range. The aim is to use AI services at most thrice a day. My current usage frequency is .. unconstrained (hopefully, the script will include some insights on this too for comparison in the results post). Okay, let’s have the grading scale:

Grading scale

5 day daily Avg AI usage	Grade
None	Super
Once	Very good
2->3 times	Good
>3	Fail

Time limits

There’s another extension I have installed called LeechBlock NG. It allows control for when certain urls can be accessed or blocked from access. I’m setting it up to allow access for 3 hours every day starting at 13:00 during which time I will have the freedom of access.

Presumptions

For this to work I will have to follow my own rules. So, let’s just hope I will. I’ll include data on how that went in the final report.

Closing

So that’s it. Just to say. Problems like these usually ain’t isolated. The AI addiction problem could be related to the Reels problem which could be connected to the YouTube binging problem .. which might be related to the music addiction problem, also to the unconstrained surfing problem. So I’m doing somework in those areas also. But I will only share about that 5 days from now. Or I should. That’s the plan. Also, I did not at any one point say AI is bad. Articles on how to use AI well are available out there. But that ain’t an area I have thought through well enough to have an opinion on. They all seem so much to me, most of them at least, like articles on how to healthily use cannabis.

I should return with the results after the challenge, 5 days later. Till then. May the force …

Oh, and we shall talk about the asterisks point soon, very soon