A Touch of The Future Shock

---

A Touch of The Future Shock

When I first started playing with AI tools back in the summer of 2021, I said I had a touch of The Future Shock. It’s only happened a few times since. Using ChatGPT for the first time in winter 2022 was another moment where I really felt the acceleration of technology.

I had that feeling again last summer when I got Llama 3.1B running locally on my machine for the first time, which also convinced me that the only acceptable price point for AI is free.

And now, over the last few days, playing with (and witnessing) two new LLM technologies—one at the high end, the other running locally—has once again left me with A Touch of Future Shock.

Operator: An AI With a Web Browser

At the high end, someone in a group chat I’m in has access to OpenAI’s new Operator tool and has been sharing their experiments.

Operator is an AI agent with full access to browsing and manipulating a web browser. It’s built using the reasoning capabilities found in 2024 models like ChatGPT-o1.

You give Operator instructions, point it at a domain, and ask it to do things—it just goes off and does them.

Obviously, Operator immediately dispatched to play Universal Paperclips. lol

As an example of the tool in action, here’s a Link to Operator cleaning up someones Dropbox folder (Twitter no longer embeds into WordPress apparently).

It’s worth mentioning that Operator is only available to OpenAI’s $200/month subscribers. $2.5K a year might seem steep for what it can do right now, but I can already see the value. The person in the group chat was using it to research AI courses on an online learning site (like Skillshare), having a swarm of Operators document the contents, then running a gap analysis to see what was missing if they were to create their own course.

It’s fascinating to watch. Just like you or I, when asked to research something, Operator will open a bunch of tabs, methodically work through them, take notes in a Google Doc, or compile information into a Google Sheet.

It sort of feels like my Menagerie of Models term. But not quite. A menagerie suggests diversity. Instead, I’d say someone using Operator has a Stable of AI agents—like a stable of horses—because each swarm uses the same model type.

Of course, this will change. I can already see the need for ‘recipes’—specialist Operators fine-tuned for specific tasks like research, gap analysis, or data entry. Swarms of AI models will be deployed from a Stable (Kennel?) to achieve particular outcomes.

Right now, Operator isn’t a “drop-in replacement for a remote knowledge worker.” It’s more like an intern whose work you still need to double-check. But we’re only three days in, and it’s going to get better.

Seeing it in action has made me truly and deeply understand the threat thatAI poses to knowledge work—especially bullshit jobs.

I have more thoughts on this, and I’m still processing. I’ll do an episode about it soon. But I’m now convinced that a lot of office jobs—especially entry-level admin, data processing, and anything that involves copying and pasting between systems—are going to disappear.

Not as soon as the AI hype beasts think, but sooner than most people realise.

Deepseek R1:32b

If OpenAI’s Operator is at the cutting edge of what’s possible at datacenter scale, Deepseek is at the other end of the spectrum.

Over the weekend, I got Deepseek R1:32b running locally on my M2 Max. It’s a Chinese inference model that was only released four days ago. Maggie Appleton wrote an short overview of its capabilities earlier today here’s the TLDR

TLDR high-quality reasoning models are getting significantly cheaper and more open-source. This means companies like Google, OpenAI, and Anthropic won’t be able to maintain a monopoly on access to fast, cheap, good quality reasoning. This is net good for everyone.

Deepseek is a reasoning model – the ‘thinking’ tech that OpenAI’s Operator above is built on top of.

R1 reaches equal or better performance on a number of major benchmarks compared to OpenAI’s o1 and Anthropic’s Claude Sonnet 3.5 but is significantly cheaper to use – either by running it locally or via lower API costs.

I have written about this twice now, and I’m going to say it again: Local LLMs like SLMs (Small Language Models) and SLAMs (Small Local Agent Models) are something people should be paying attention to.

It’s crazy that I can have a reasoning model this good running on my own machine!

I used Ollama to install Deepseek R1:32b and found it burning a hot 20GB of RAM.

I gave it a section of a talk I was struggling with and asked it to summarize the key themes. Then, I asked for editorial advice on a rewrite. The feedback was solid—better than anything I’ve received from previous local LLaMA models.

Interestingly, at the end of its response, it suddenly switched to Chinese – which isn’t surprising given that it’s a Chinese built model. (I joked on social media that my idea was too powerful to express in English.)

Then something weird happened. At the end of its response, it suddenly switched to Chinese. (Not surprising, since it’s a Chinese-built model. I joked on social media that my idea was too powerful to express in English.)

Google Translate told me the sentence ended with: “entertainment and cultural exchange, thus affecting the way the entire society interacts.”

I asked Deepseek itself to translate the same line, and it gave me basically the same result.

A local model, running on my laptop, just produced the same translation as Google Translate. Sure, it’s burning 20GB of RAM, but further distillations and optimisations will come within weeks, not months.

As I said in my review of Apple Intelligence more and more capabilities are going to get squeeze into local models running on hardware you all read own. Apple Intelligence is ‘fine’ right now, but its going to get way better.

Exploring both Operator and Deepseek together—new tools released in the same week—is why I’ve got A Touch of the Future Shock. (Though it might also just be my head cold and high temperature, in which case, it might just be a touch of the flu instead.)

Remember: Right now AI is the worst it’s ever going to be. 

---

On The Blog:

Issue #12 of my zine went out to mailing list subscribers this week!

Start Select Reset Zine: Shadow Boxing

Issue #012 is a double feature, tackling two intertwined creative struggles:

• Shadow Boxing – The exhausting mental habit of preemptively battling imaginary critics online.

• Long Form Leap – My ongoing attempt to write longer, more ambitious essays.

If you would like one too consider subscribing to my mailing list! I have a couple of back issues of #11 left too.

I also made several changes/updates around the website this week:

• I continue to get old episodes of Permanently Moved on to the blog. We are now very deep in to the archives from 2019.
• I updated Solarpunk category the landing page. Tided up some text and layout – lots more to do on here.
• Refreshed my Support Page copy and made some notes about re-building it entirely

---

Photo 365

The Ministry Of My Own Labour

• Plugging away at a talk for Friday
• Worked on Project ENTRY
• Had a great call with someone at an AI Eval lab
• Felt ill since Friday morning 🙁

Terminal Access

via Paul Graham Raven in my DMs comes this long read: The Last Boys at the Beginning of History.

The piece examines the emergence of a new generation of young, predominantly male intellectuals whose political consciousness was shaped during Donald Trump’s rise to power and his first term. Unlike earlier conservative thinkers, these individuals lack memories of belonging to any political or cultural milieu other than Trump’s.

At the last afterparty in mid-December, I meet a distinctly new kind of conservative. “I’ve done ayahuasca eleven times,” he tells me. He is close to thirty, I think, and has been introduced to me as a great scholar of the Bronze Age, or so I hear with excitement—it turns out he’s an expert on Bronze Age Pervert. “Each time I’ve gone to the ritual in the Amazon, I’ve brought Thus Spoke Zarathustra there and read it.”

“During?”

“No, not during,” he says, as though I ought to know how one typically reads Nietzsche’s tome at an Amazonian ayahuasca ritual. “Before and after. Chapters at a time.”

“The last time I did, it”—the ayahuasca, he means—“told me to never return. And I haven’t since.” He pauses and looks thoughtfully away from me; he sighs. “I came away from it certain that modernity must be destroyed.”

The article present a real challenge to the liberal portrayal of Gen Z Trump supporters as merely ‘apolitical’ or easily swayed by podcast media. Instead the article highlights their genuine ideological engagement. A really fascinating look at the evolving landscape of conservative thought makes me think about the potential implications for future political dynamics in the US.

Reading

I’m re-reading Autobiography of a Yogi by Paramahansa Yogananda. In the intervening time since I last read it i’ve done like .. 5000 hours of meditation and the sentences/wisdom that jump out at me this time around are very very different from the first time I read it. This book is a total masterpiece.

Still reading How to Take Smart Notes by Sönke Ahrens. I’ve made some changes to the way I keep and take notes already based on this book, and I look forward to reading more.

I’m chipping away at Immediacy: Or, The Style of Too Late Capitalism by Anna Kornbluh. This book is so dense, and I’m really struggling with the prose/writing style. Trying to get though a section a night. Something like 1-2% of the ebook at each sitting.

I ditched reading Immediacy as soon as I started to feel ill on Friday and spent much of my time moping in bed reading book #13 of the Gaunt’s Ghosts : Salvations Reach by Dan Abnett

Rebecca Black – Sugar Water Cyanide

The new single from Rebecca Black is a banger! So good in fact I’ve already heard it being played in my local coffeeshop!

It’s a little strange to me how all new singles now sound like 4 songs jammed together, in order to best optimise for short vertical video media ecosystem but it really works in this track.

Remember Kids:

The Universe is worked and guided from within outwards

H. P. Blavatsky – The Secret Doctrine