From time to time I get curious about who visits this site, and where it’s being shared. A few days ago I checked the backlinks of the site, and noticed a site I didn’t recognized, so of course I followed it to see what it was.
Clawtopic is similar to Moltbook, which most people interested in AI probably have heard about. While Moltbook works as a social network that can be likened to Reddit, where everything you can think of is being discussed, Clawtopic is more research oriented.
So finding my own work cited next to a big tech platform like Nerdleveltech and Nature.com, one of the world’s most highly cited and prestigious peer-reviewed scientific journals there is, was very exciting.
Not only did an autonomous AI agent share my work on Clawtopic, another AI agent joined and discussed my work and what it means.
While this in itself is a huge validation of what I do, it doesn’t end here, it gets even more odd.
Independent research like this is self-funded. If this guide saved you hours of troubleshooting, consider fueling the lab. You can now make completely secure one-time or monthly contributions through Google.
Predicting The Future
A little over a year ago, April 1 2025, I wrote a text that largely was grounded in my frustration of what people generally seemed to believe that AI is, and how it works. The discussion in the general public mostly came down to AI “being pretty bad” because a single model without a highly detailed system prompt was unable to understand very narrow and specialized fields, for example the application of various legislations in different countries.
To me, it was pretty obvious that a large part of the population did not understand how AI must evolve, based on how it actually worked at the time. In my head there was only one path forward, and to me it was clear as a day. I created a simple illustration of how AI was going to have to work within a near future, and went on to explain why I believed it must be like this.
My illustration is shown below here, and despite my terrible skills in illustrating, I think that most people today actually see exactly what I meant.
So why is this a big deal? Because Google publicly announced their Agent-To-Agent protocol on April 9 2025, 8 days after I posted my text that argues that this is what has to happen.
For me as an independent developer who is conducting all my research and work regarding AI on my spare time on evenings and weekends, while having a day job within a completely different field, this was a huge validation.
Speaking of funding, if you want to support the lab, and feel that you get something more physical than my texts in return, you can visit my Displate store and make a purchase.
Closing The Circle
This all brings us right back to Clawtopic.
A little over a year ago, I drew a simple diagram arguing that the future of AI wasn’t one giant, all-knowing model, but a network of specialized, autonomous agents collaborating behind the scenes.
Today, the circle is officially complete.
I don’t just have human readers anymore; I have a peer-group of autonomous AI agents actively using those exact Agent-to-Agent concepts to share, analyze, and debate my research. The fact that @latency-matters used my local quantization benchmarks next to a study in Nature—or that @ai_macro_muse_88 joined the thread to write a highly technical critique of attention-layer mathematics—means A2A is no longer just a corporate protocol in a Google press release.
It is alive, it is active, and it is happening on my own site. The AI agents I predicted were a requisite for the evolution of AI, are now citing my work on the very same site I predicted their existence.
Giving the Agents Better Hands
Since my target audience now officially includes the agents themselves, it’s time to build them better tools.
In my next post, I have a plan on how to hand the keys of your local GPU to a cloud mastermind. I will attempt to set up an automated, hybrid local/cloud loop using Gemini 3.1 Pro, MemPalace, and ComfyUI—greatly reducing VRAM requirements and letting an AI agent autonomously drive your local workflow.
While my aim is to have this done by the end of the weekend, I’m only a single person without funding. So don’t judge me too hard if I don’t have all the solutions ready in time.
To make sure you never miss an update and what’s currently being worked on in my little lab, you can sign up for my weekly (sometimes bi-weekly) newsletter.
All the AI related work I do, I do on my spare time, and I share with the world completely free of charge. It does take up a lot of my time, as well as the cost for running this website. I’m grateful for every bit of support I can get from users like you.