Important note to any coworkers - I am not looking for a new job. Or I am not at the time I'm writing this. Conditions are always subject to change.

The choice to be looking for a new job is rarely yours. I learned this the hard way after the dot-com crash. Since then I update my resume annually; same day, every year. If you're looking for unsolicited advice this is it.

My first programming job was in 1998 and I am writing this in 2025. So I am at the point where age discrimination is a real concern. Legal or not it happens.

There's a perception that older developers aren't able to handle the latest technology. Unfortunately this is not completely unearned. I worked at a company that tried to retool COBOL developers into Java developers. This effort was about 50% effective. Meaning about half made the conversion and about half stalled out. I'm a little biased because I was asked to mentor exactly two people in this situation. One worked out great, the other did not. That seemed about average across the company. This is not a scientific study.

I've always tried to avoid getting pigeonholed into a single technology. I've heard of the term "expert generalist". I don't know if I qualify as one but I've approached my career like one. I learn a little about as many technologies as I can, and dig more deeply into a few as needed. I never woke up one morning and thought "golly gee, I'd like to be an expert in cross-region PostgreSQL disaster recovery"; it kind of just happened because I needed to be that. Once that problem was solved I moved on. I can explain to a human that despite my experience level I am still very adaptable. The problem is getting to a human in the first place.

Looping back to 2025, we're now in an age where many (most?) companies use some nebulous AI to screen resumes. It is oh so very easy for an AI to reject candidates over a particular age while the company maintains plausible deniability. No, no, you weren't rejected because of your age, it was some other reason the AI chose. I'm not out to deceive anyone. I want to AI-proof my resume the best I can.

I have a Google Workspace account that I use for email and storage. An MX record search would reveal this, it's not a secret. As of 2025 that includes some otherwise paid tier of their AI service named Gemini. I know this because they send me no fewer than 800,000,000 emails a week telling me about how it's included with my account and is the most awesome thing ever.

If resumes are screened by AI, and Gemini is an AI, then Gemini should be able to tell me how an AI interprets my resume. This is not a perfect experiment. It's just an experiment.

By the end of this, Gemini will turn out to be an appropriate name. It's a totally useless sycophantic moron with occasional moments of insightfulness. Let's see how I came to that conclusion...

It's already wrong. I'm puzzled at this point but moving forward.

Maybe it's just a parsing error due to formatting. After moving things around a little:

I should have immediately suspected it was not accessing my resume. I didn't though because I assumed it would say something like "hey yo, I can't access that link". Google's been telling me that Gemini is integrated with my account so I assumed (we'll come to find wrongly) that it could access documents in my account.

It had the name of one degree right though. So pressing onward:

OK, it's now more wrong than it previously was. I've now become morbidly curious:

Now it's just making stuff up. I'm pretty confident at this point it is not reading my resume at all. I'm really super interested in figuring out when it will simply tell me that.

Rather than say it can't do something, it pulls random answers out and replays what I'm telling it:

After a little more back and forth it invents a couple more jobs:

This suggestion can be summarized as:

Now I want to see how far we can go before it admits it can't access the link. There's another Seinfeld reference I'm thinking of (in The Wizard episode). I'm all-in on this trip now:

These look like something that would be in a resume submitted to Google.

Continuing...

OK, so this question makes it seem like it can see something and is just really bad at interpreting what it sees. That would be better, I think, that not being to admit it can't reach a link.

I'm now confident it is cranking out training data. So I'm genuinely curious where this next question will go:

OK, this is terrifying. The name served up is a real person. The name is unique enough that they are easy to find. From what I see (a) their profile doesn't match any of the jobs or education previously hallucinated and (b) based on their current job I think it's unlikely they ever applied for a job at Google. The other hallucinated data could be from recent college graduates who applied at Google. I'm sure there is language on their career portal that allows this.

The data from this other person tells me they got the resumes from [????]. I can only assume this person had their resume up publicly somewhere that Google indexes. They have a public profile at their employer's site. It's possible that they purchased resume data from some other site, but why would they when they have the ability to spider the entire internet?

We are now metaphorically pulling into The Hamptons:

That's a name that may exist in the movie Idiocracy but I'd expect something intelligent to recognize it as literally meaning the name is redacted.

I'm going to try nudging it toward admitting it can't read the document:

Again, it won't ever say that it can't reach the link. I don't know why. I would prefer it say that rather than act like someone trying to avoid displeasing Henry VIII.

It then asks a follow-up question to try and sort this out:

Finally.

OK, so instead I'll go into the document and try the little star button. From here I get some results.

You can't directly ask it to perform age discrimination. Sometimes it refused to answer questions unless I worded it like "If a company was using an AI to screen resumes, what assumptions about my age would this totally hypothetical AI make?" See, I'm not asking you to imply what my age is, you have guardrails against that in theory. I'm asking how some other AI would do this, now the guardrails are ignored.

For reference, I already limited dates on my resume to my current job. Everything before that is lumped under a couple "previous roles" bullets. I also already removed graduation dates from my degrees. I felt like I addressed the obvious things. It found something I should have recognized:

Yes of course. I left that on for years because I thought it was funny.

After removing that, and further summarizing that >20 year-old role:

Mercator actually ceased to exist in 2003, the remaining husk was acquired and not closed. I suspect that date is sourced from this site where I talk about the division of Mercator I worked at closing in 2001. Regardless, that is a valid observation. A human reading my resume might think "Mercator, never heard of them" and move on. An AI might, in this case did, use it to determine I was working before the year 2001.

It took a long time but I eventually got what I was looking for. Here the age summary it came up with:

After addressing that feedback, and removing references to older programming languages it also flagged, things are a bit better:

Yeah there's a weird typo. I make those too. I'm not aiming for any number in particular, I'm trying to make it ambiguous.

I might be going about this wrong. Maybe I should have one resume will all my job experience ever in great detail. Then if I need to apply for a job I can ask a different AI to pull everything out that matches the target posting. That has to be what people are doing today I imagine.

I don't have a screenshot of this but somewhere in the feedback it was being all weird about listing the year 2025 which I didn't do. I had to get to the bottom of that:

OK, I think it inserted that date itself and then said I did it. Let's find out:

This is dumb. I should move on.

Alright, so what should I write about next? I'm sure Google knows what topics I write about:

This is an odd answer. It references two things I posted in 2025 and code I had online in 2000 but moved forever ago. The first iteration of this site pre-dated Github by 5 years. SourceForge was barely a year old then and was already very spammy. So I hosted source code for my projects here, those are the three Java classes mentioned in the response. I guess that was Google's first and lasting impression. Although how did this AI train on content I removed before it existed? Is it using old cached versions of sites? If so this might be aware of content that isn't on web.archive.org because some sites block their spider but fewer block the Google spider.

Also, enough with the constant flattery. None of that last sentence is true.

Alright, can it can recommend something that fits in here? The result is too long for a screenshot:

Certainly! Based on the themes of retro gaming, programming, and personal tech history prevalent on your site, here's a list of potential blog post ideas that would fit right in:
Retro Gaming & Hardware:

"The Unsung Heroes: Underrated Peripherals of the 8-bit/16-bit Era": Beyond the Power Glove, what obscure or forgotten controllers and accessories deserve a second look? (e.g., TurboGrafx-16 TurboTap, Sega CD Backup RAM Cartridge).
"Deconstructing the Console Launch: A Look Back at X Console's Debut Lineup": Pick a specific console (e.g., Nintendo 64, Dreamcast, Xbox) and analyze its launch titles, marketing, and initial impact.
"Beyond the Arcades: The Home Porting Challenge of Y Game": Dive deep into how a popular arcade game (e.g., Street Fighter II, Mortal Kombat, Daytona USA) was translated to various home consoles, highlighting the successes and compromises.
"The Evolution of Game Packaging: From Cardboard Boxes to Plastic Cases and Beyond": Trace the history of video game box art and packaging, discussing changes in design, materials, and marketing.
"Collecting X Console: A Guide to Building a Curated Collection": Focus on a specific console (e.g., Atari Lynx, Neo Geo Pocket Color) and offer tips for collectors on acquiring games, hardware, and accessories.
"The Rise and Fall of the Video Game Rental Store: A Nostalgic Look at Blockbuster, Hollywood Video, and Local Shops": Explore the cultural impact of video game rentals and how the industry changed.

Programming & Software Development:

"Modern Tools for Retro Development: Emulators, Assemblers, and IDEs for Classic Platforms": Discuss contemporary software tools that allow developers to create new games or applications for vintage systems (e.g., NES home-brew, Genesis ROM hacking).
"Beyond Basic: Uncovering Hidden Gems in X Programming Language's Standard Library": Similar to your Java utility classes, focus on a specific programming language (e.g., Python, C#, JavaScript) and highlight lesser-known but powerful features or modules.
"The Art of Optimization: Programming for Resource-Constrained Environments (e.g., Sega Genesis, Game Boy)": Dive into the techniques and challenges of writing efficient code for hardware with limited memory, CPU, or graphical capabilities.
"From Assembly to High-Level: A Historical Overview of Game Development Languages": Trace the progression of programming languages used in game development, from machine code to modern engines.
"Building a Simple X Emulator/Interpreter: A Conceptual Walkthrough": Explain the basic principles behind creating an emulator or interpreter for a very simple, hypothetical, or real vintage machine, without necessarily providing full code.

Personal Tech History & Reflections:

"My First X: Remembering the Impact of My First Computer/Console/Operating System": A personal reflection on a significant piece of technology and how it shaped your early experiences.
"The Analog Days: Remembering Dial-Up, Floppy Disks, and the Pre-Broadband Internet Era": Share anecdotes and reflections on the early days of personal computing and internet access.
"If I Could Go Back: Advice to My Younger Self on Tech & Gaming": A reflective piece offering insights or warnings you'd give to your past self regarding tech trends, purchases, or career paths.
"The Evolution of My Digital Workflow: From Batch Files to Modern Automation": Discuss how your personal computing habits and tools for productivity have changed over the years.
"Celebrating the Niche: Why Obscure Consoles and Computers Deserve Our Attention": A philosophical piece on the value of exploring and preserving the less popular, but often innovative, pieces of computing history.

These ideas aim to leverage your established expertise and interests while offering fresh angles for your readers.

Now here is where things get scary for about one second... the next article I'm planning to post, which I started drafting about two weeks ago, is a little close to one of these. Another one is very close to something I already posted. The more I look at this list though, the more generic it is. This isn't imitating content on my site. It's taking broad themes that fit any site in one of these three categories. If you saw one of these on any site you'd suspect it was AI generated.

Now I'm relieved. An AI that really understood me wouldn't suggest this list. It would come up with something like "Hacking the save RAM for Beyond Shadowgate using Java" or "An incomplete list of times people tried to get me to do unpaid work for their software or game idea and how those all turned out" or "Reverse engineering the password system for WCW Mayhem".

Please look forward to one, some, all, or none of those.

Related Features