Alien Tools and Answering Machines

Reflecting on 2025

2025 was a much different year for me than most.

This year was the year I felt I really hit my stride at work. I’m a developer advocate at Pinecone, and a lot of my work is intensely public, so it’s easy to feel a strong sense of imposter syndrome around literally anything I produce. This year was my second year in the role, and I got to do a LOT of cool stuff:

• I wrote a few really solid pieces for the Pinecone Blog, such as this piece on sparse retrieval and this piece on context engineering.
• I made a tooooon of videos for our Youtube channel, which resulted in hundreds of thousands of impressions and tens of thousands of views
• I gave huge webinars and talks on platforms like Analytics Vidhya, AI Tinkerers, and AWS Re:Invent
• I ended the year publishing a plugin for Pinecone in Claude Code, which took off and now has over two thousand downloads!

I finally feel that I am good at something, or at least, clearly getting a little bit better at this something. I owe this to my team at Pinecone and many mentors I’ve accumulated thus far. Thank you!

It was a good year personally too. I went on lots of trips, spent quality time with my friends and family, worked really hard on my surprise engagement to my beautiful fiancée Divya, went to soo many weddings, ate so much good food, vibe coded a bunch of fun stuff and managed through all the geopolitical nightmare fuel of it all too!

The World and Tech Change

I started my career in tech, but I really matured during the rise of deep learning and AI. You can see me try my hardest to sound like I know what I’m talking about here in 2022, when I won my first hackathon for Cohere using my simple prompt engineering demo, Perfect Prompt. By the way, that was back when pioneering LLMs were just prompt completion endpoints, rather than instruction-based… Oh, how time flies and how fast tech grows!

AI has gotten so interesting, so useful, so unusual, a little scary, and really weird. The development of AI tech has progressed so quickly, I feel like I’m constantly trying to stay afloat. I mean, let’s take a moment to acknowledge that so much of the modern tech industry is dominated by large language models. That even the majority of the code that I write myself is now facilitated through a large language model!

There’s so much to know and not know, and I often think of Karpathy’s descriptions of LLMs as “alien tools”, objects that perform with superhuman ability and incomprehensible scale and performance.

Hearing his experience (and tons of other from developers) excites me, but mostly because I think there’s a lot of great boring stuff that can be done with a lot of benefit to humanity, and some fun stuff too. But some people don’t feel this way. There’s a strong sentiment of fear and uncertainty and doubt all around AI, and for good reason. Why wouldn’t you, among all this talk of job loss, layoffs, and potential for infinite content generation and misinformation, feel concerned?

Look, I love a sci-fi metaphor as much as the next techie, but describing these things as alien tools isn’t doing anyone favors. Alien tools are unpredictable, menacing, and mysterious. They resist methods for understanding and often endanger their subjects.

I mean, they do seem foreign, insofar as the ocean or deep space is foreign and mostly unknown, but not unintelligible. They are indeed objects that deserve more study, but maybe not inside a hermetically sealed chamber.

I like to think about LLMs as answering machines rather than alien tools.

A very old-school answering machine from the Smithsonian

Answering machines were once used to help companies that couldn’t afford full-time secretaries keep track of people calling in requests (source). Then, they were widely adopted since sending and saving messages was so convenient for people.

Modern answering machines have similar, but more literal capabilities; “ring” one up, and expect to leave an information-rich correspondence.

Instead of waiting for someone to respond, you get an answer pretty instantaneously. The quality of this answer depends on the bearer’s message and the ability for the machine to retrieve contextual information, and there’s only so much information you can communicate.

And a good answering machine has a lot of engineering to ensure that a quality answer can be provided. A look inside the lids of Claude or ChatGPT would reveal these services to be highly architected tools.

Of course it turns out that even the original answering machines, while seemingly simple, left a huge impact on society, transforming social networks and the nature of telephone communication itself, not always in good ways, either in good ways either.

Everything's old is new again, isn't it?

Even with all of that deliberate design, it seems to me we’ve built these answering machines as if we already know how to use them, but it turns out we haven’t quite figured out how to ask the right questions yet.

And maybe if we think of them as answering machines rather than alien tools, we’d be more willing to roll up our sleeves and make these things better for our world, while tackling the extremely thorny problems they can create too.

Let's roll up our sleeves and learn some stuff!

While I don’t think I can resolve this tension head on (maybe one day!), I can at least do something I’m really good at. Make things that are hard to understand, a bit easier, with some humor, reflection and commentary. Write, record, or make things that make AI a little more interesting and approachable, and maybe that’ll help someone do a thing they wanted to, or at least, learn about what is actually happening with modern AI without needing to read a textbook or three.

So I present to you: Answering Machines, my new blog and newsletter on all things modern tech, AI, the world, random stuff, and yours truly. It’ll be infrequent, irreverent and probably a little too much. I’ll make videos sometimes, at least one article each time I send an email, and point out anything I find interesting. Oh, and I do a lot of answering-to-machines nowadays, so I hope to have something to share about that part of the job too.

There’s so much rigidity and corporate-speak when it comes to AI and tech. I want to create a tiny space on the internet where I can talk about what to really make of developments in AI, and how that relates to ourselves and the world.

So if that’s something that you think will be fun to read too, go ahead, sign up, and if you’re feeling up to it, leave me a message after the beep.