Playing AI Far Cry

My AI Learning Strategy based on Playing Far Cry

There was a brief, glorious time when self-taught technologists could compete in the job market. Yes, there is (or was—the job market changes so fast now) a push by some companies to hire developers without degrees, but the intent seems more about paying less for the same (or better) results rather than valuing capability over pedigree.

To learn on your own works best with intrinsic motivations, which can be quite diverse. Some were looking for a brass ring or golden ticket (both Boomer terms kept alive by boomers running Hollywood) to trade their way out of a dead end job or financial poverty. Some were just curious, while others just had a knack for tech and enjoyed the challenge. Or, in my case, all of the above, though it started with playing PC games. I had a low-end, second-hand Packard Bell and a BBS account (look it up ??). Access to shareware games (look that up, too ??) was only half the battle for someone on a $0 budget.

Many game developers target their code at the latest and greatest hardware and not a three-year-old department store discount model with ten more payments to go. I learned to tweak system files to boost performance enough to play the games. I became hooked enough on the games to write game reviews for a couple of now-defunct ezines in exchange for a byline and free games. It also revealed that a PC was more than a platform for gaming and desktop publishing and could be used in god mode!

One of the last games I reviewed before becoming a paying consumer was Far Cry 2. It is an open-world game, meaning you can basically do anything you want and go wherever you want, so long as you can survive or access it. It has an interactive plot, where choices made at certain points will affect later options and opportunities. This wasn’t a new approach; Wing Commander had been doing it for years, even featuring plotlines as cut scenes with well-known actors. But it was the first time I had seen it in a first-person shooter, my personal favorite genre because of the immersive experience (and the popularity of Doom shows I’m not alone in this). Acknowledging both Far Cry 2 and the Wing Commander franchise as a whole, the metaphor continues with the next chapter in the Far Cry series: Far Cry 3.

Far Cry 3 introduced acquiring skills, knowledge, and ingredients. You learned not only how to aim a gun but how to steady it for accuracy. Ingredients, in the form of animal pelts and plants, could be accumulated, and then you learned how to make things with them by combining them in specific quantities (called crafting). The first time I played through, I would just try to accumulate and improve weapons or craft upgrades while running through the map to conquer everything and finish a plotline along the way. It wasn’t until very near the end that the power of skills—and the value of applying the right skills to the right situation—became clear.

In every game since, I start by avoiding the plot and pursuing skills. I’ll step into the plot at points when it’s necessary to acquire a skill or sometimes an ingredient, and then go back to building my capabilities. Once I’ve taken that as far as possible, I enter the plot. This strategy leads to some games where I make it from start to finish without a single death (admittedly, that happens only about 2% of the time).

I’ve lately begun to tell people my superpower is digression. I think I’ve just demonstrated that once again.

What does this have to do with how I’m learning AI? While all the prognosticators are talking about how AI will do all our jobs soon, I’ve been too busy doing my own job to spend as much hands-on time with AI as I’d like. But I do have pockets of time to read articles, listen to podcasts, and watch YouTube until I can get hands-on (which should be very soon). It dawned on me the other day that the concepts from articles and podcasts are like collecting recipes and ingredients for later crafting—stuffing them into larger rucksacks and adding tools as I go. Meanwhile, my expanding YouTube subscriptions and playlists are like gathering more accurate and powerful weapons for taking on the more challenging enemy forts, such as building local LLMs, coordinating MCP implementations for private GPTs, and managing secure agentic operations.

I make most of my playlists public on my YouTube channel. There’s a collection of AI Learning playlists there (it’s the second collection down) that will continue to grow. I’ll also be sharing my discoveries on my blogMedium, and Substack. New posts will be announced on my social accounts, all @scottsnelson1. You won’t find me posting on Instagram yet, because my daughter-the-influencer still hasn’t had time to teach me how to create posts with links (queue Harry Chapin), nor on Pinterest because it’s more work than I have time for right now. And, of course, one of my first MCP solutions should be to manage these posts for me. I’ll let you know when I defeat that fort on the third map.

(Alternate close written by Perplexity.ai)

If you’re an IT stakeholder or technologist, you know that learning new tools and frameworks is a lot like leveling up in a game. You collect knowledge (ingredients), build skills (crafting), and tackle bigger challenges (forts) as you go. I’d love to hear how you approach your own learning journey—drop a comment or connect with me on my channels. Let’s build our skills together and take on the next big challenge in AI and IT!

If you found this interesting, please share.

© Scott S. Nelson
Agentic Orchestration

Quick Thoughts on Orchestrating AI Agents

Revising some thoughts from a recent chat with a longtime friend…

In agentic AI, it helps to think of it more as declarative programming than as prompting. While agents rely on prompts, they also perform multiple operations and follow a declared path.

There are also architectural considerations for organizing and orchestrating agents by purpose, capability, access, trust, and cost.

But that’s just my opinion. I could be wrong” – Dennis Miller

If you found this interesting, please share.

© Scott S. Nelson

TIL a Cool PowerPoint Designer Hack

I don’t consider myself creative or artistic. The artistic bias probably goes back to school, where the requirement of an art elective ruined my chances at a 4.0 grade average (and exemplified by my Featured Image attempts, which even AI can’t seem to help me with).

The creative bias is that I usually can’t do it on demand. I’ve also learned that if you get out of your own way, creativity happens fairly naturally. This is similar to how Douglas Adams describes how one learns to fly. Anyway, one way to get over a lack of creativity and artistic sense is to let AI tools do it for you, and one that I find really handy for my handicap is PowerPoint Designer. (For those who think Designer isn’t AI, marketing is clearly doing it’s job, and perhaps missing the point, and here is proof— as long as the link is valid).

Personally, I don’t mind boring slides. I actually like have really basic slides where the purpose of the slide is for people to remember what we’re talking about when their minds inevitably wander. But, sometimes I am creating slides for other people, or need to meet the expectations of people with different opinions on the matter, so I need to make them a bit less boring, and designer is a great way to do that. One frustration I have with designer is that it so often gives me this message: Microsoft PowerPoint Designer "Sorry" message.

Usually this can be fixed by simplifying the slide, i.e., remove the cool graphic you added and let it focus on the text. Or you can split the graphic and text into separate slides and then combine the results after the magic happens.

Oh, and one word of caution: Copy your original slide before letting Designer have its way with it, because sometimes the new formatting is no longer easy to copy and paste.

Back to the cool hack part. My second annoyance with Designer, after it apologizing for having no suggestion of how to improve my work that clearly needs improvement, is that it gives so few suggestions. This seems to have gotten worst, and I suspect it is because those data centers are saving cycles for the AI that the marketing folks are calling AI. Recently, it was only giving me four or five options, many of which were just minot variations on the themes, like this:

PowerPoint suggestions for a slide with plain bullets.
PowerPoint suggestions for a slide with plain bullets.

I wasn’t too thrilled with any of the options, but I picked one just to move on and make some progress. Maybe an AI image would spruce it up enough (in the end, it did). Being the paranoid person who has lost early versions that I wished I could go back to, I made a copy first. Usually, when I make a copy, I start working in the copy. But this time, for no particular reason, I went back to the original…where Designer was showing entirely new options based on it’s own modification:

PowerPoint Designer suggestions after accepting a suggestion.
PowerPoint Designer suggestions after accepting a suggestion.

It seemed I had accidently cracked the code to get more options, like in the good ole days before everyone was using these tools, too. Just to prove my theory, I tried repeating the process, and sure enough…

PowerPoint Designer Keeps on Designing
PowerPoint Designer Keeps on Designing

I didn’t really find  an end to variations, though I admit that the quality of options generally declined, with an occasional interesting one coming up here and there. Full disclosure: this may have just been the raw material I started with, but that is back to my bias against my own creativity.

So, that’s my big discovery for the day. Well, there were really more, but I have to get back to “real work”, until this writing stuff actually starts paying some bills. Forward this to your friends (or enemies) if you would like to contribute to this hobby.

If you found this interesting, please share.

© Scott S. Nelson