What even is an “AI Time Capsule”
Well, it’s not a fancy new library, a company built around a prompt, or an interesting way to interact with AI. This is just going to be a post about my thoughts on AI as it currently exists today. As of the first draft of this post, it is April 24th, 2026. The motivation behind this post is to catalogue my complex feelings about AI so I can look back and see how they have evolved. I don’t think I’m going to say anything that hasn’t been said before. For this post, when I refer to AI, I’m speaking about generative AI.
What’s going on in the world of AI as of writing?
There is so much news across so many fronts that frankly, it’s exhausting. Tech has always moved fast; that is one of the reasons I’ve always loved it, but this is another level. Each day we have a new model, a new harness, an earth-shattering breakthrough, a new product to learn, a new product getting cancelled, and so on…
That being said, OpenAI released GPT-5.5 yesterday. A little over a week ago, Anthropic launched Opus 4.7 to an incredibly lukewarm reception alongside announcing Project Glasswing because their new model Mythos is so powerful that we mortals couldn’t possibly handle it, it’s not a marketing ploy at all.
In the open source world, Deepseek announced v4 of their API (no, I couldn’t find an official announcement anywhere other than Twitter, and I refuse to link to that site from here).
Qwen released Qwen3.6-27B, which is a glimmer of hope for open source models that users can actually run locally. Google released Gemma 4, with some incredibly cool methods for getting the model size down so low it can run on an iPhone. These days, most of my hope for AI exists in these open-source models that we as users can hopefully run locally one day.
Were tight on tokens
One other trend I’m seeing across the industry is companies raising costs for tokens. GitHub is pausing new signups for Copilot Pro, Pro+, as well as removing access to Opus models for existing users. Anthropic “A/B tested” removing access to ClaudeCode from their pro subscription (link to the thread on HackerNews since I refuse to give Twitter any more traffic). In the enterprise world, I use both Cursor and Langdock at work. Both have recently implemented new usage caps/pricing schemes that drastically raised the cost to use both those tools. In some cases, the bills have increased more than 70x. Is this a “rent is due” moment? Who knows, but if the trend continues, I wonder how much longer smaller shops can justify prices like this.
I’m tired, boss
All of these AI companies have these generative AI models and are rushing to find a way to monetize. As I mentioned in the intro, these companies are quickly jumping on any chance they can to shove their model into something new. Off the top of my head, in the last few weeks, we saw Claude Design, OpenAI doing images again, as well as trying to desperately get some that enterprise cash by shoving ChatGPT into Excel.
That’s just the frontier labs, don’t even get me started on what everyone else is doing with these models. Every website and service under the sun has an agentic first step for every customer service that just regurgitates help docs to you. Companies are mandated to put something to “… for the AI era” in their taglines so we know they are hip. Are we also putting AI in toilets, I guess? Keeping up with everything is tiring enough; it’s adding friction to interactions with companies and services I use. I do not want AI in everything I use, especially not enabled by default. Am I out of touch? Do people actually use these services, or is it strictly performative? I’m tired of each new marketing stunt, each new claim that “AGI” is here, all the new workflows and tools and magic prompts. I’m just over it.
The human of it all
It’s possible that a discerning reading picked up on it, but…I do not like AI. Especially as it’s being marketed at and sold to us today. I do not believe the billionaires selling this to us have our best interests in mind. They don’t want AI to usher us into an era of universal income, where humans are unburdened from knowledge work, and the creative human spirit can endure. Instead, I think we are seeing it used primarily as a scapegoat for massive layoffs(two other 10%+ layoffs occurred in the same week as this one, also blamed on AI). The future that seems most likely is more akin to the Warhammer 40k servo skull than something more Star Trek-esque.
It seems increasingly likely that the promise of the white-collar worker is fading. AI, even if it isn’t actually taking jobs from entry-level workers, is already being used to justify the slowdown in hiring. Without juniors, who is going to become the next seniors? Are we just pulling the ladder up after us? I used to enthusiastically recommend this career to younger folks; I no longer do. I still tell people to learn how to code, but the job I loved has already been altered so drastically, and I’m not sure that it will ever be the same again.
AI is also reducing human interaction. Increasingly, I’m getting release notes or meeting summaries that are pages long, where in the past it was a solid bulleted list. People are clearly responding to feedback or questions by asking ChatGPT. My blood pressure rises each time I read “ChatGPT said this…” or “I asked AI, and it said” It’d be more useful to just send me the prompt they used rather than copy and paste a blog of knowledge I have to parse. The point is beautifully illustrated in this blog post.
So what…
I’m not sure where to end this. I think it’ll be interesting to revisit the post in six months or a year and see what’s changed. In five years, will I look back, with my self-hosted on-prem AI, laughing about this post, or will I be unable to afford anything that could even resemble a graphics card or RAM?