Gavin Owens

Gavin Owens

@gavinowens.net

founder @ https://tonk.xyz 🐞 @ https://termite.works personal @ https://gavinowens.net small software, cooperative computing, tools for thought, tech governance and free media

157 followers61 following112 posts

I am a very earnest person (American) and serious poster, but since it’s Sunday here is a picture of me and Ulriich, our house pigeon:

Whenever I’m feeling down or envious I just remember that so little is needed to max out the human experience; I already have enough and that is the best any human being can do. The rest is for the love of the game, so go out and enjoy. A little reminder for you all today.

@neurocy.bsky.social showing off hazel. Fun fact, the IDE we built at my old company Deco had some similar features like inline elements for data types, like sliders, but it was done in such a dumb way that hazel makes me ashamed and embarrassed because it is very impressive.

I will release the lecture we held on digital micropublics as a tonk, but first we must support streaming because the recording is 4GB 😱 it’s in the works!

I have a recording of the lecture and discussion with a 50min interlude clip of Guattari interviewing Tetsuo Kagawa. Where to release it?

What is a digital micro-public? That is the main question of today’s upcoming event. We will broadcast live from the Termite Undermound in London today at 2pm BST. Tune in @ tinyrad.io, going to be very special.

Kind of mind blowing that this would work and actually be really cool.

Eileen (╹◡╹)@bumble.blue

My dream open and local setup for playing with computers: @pi.dev as the harness, @tonk.xyz as the database, @atproto.com for identity, and @iroh.computer for sync 👾👾👾👾

I've been prototyping a new playing deck in Tonk for @getdweb.net . Here's the browser and edit tab. I make a lot of changes through my agent (it's useful for pulling information), but obviously I need to go in and tweak the copy or correct it in places.

Gavin Owens reposted

A small ask from the local-first community. I’m speaking at @localfirstconf.com in Berlin about @tiles.run, but this year the conference isn't able to cover speaker travel. I'm trying to raise $700 through Github Sponsors to cover my flight. Any amount helps. github.com/sponsors/til...

github.comSponsor @tilesprivacy on GitHub SponsorsWe believe that privacy adoption at scale must work backwards from preserving utility. We aim to deliver “just works” experiences through best-in-class privacy-focused engineering while offering un...

There is a twisted irony for companies right now. In Satya's latest article "A frontier without an ecosystem is not stable," he warned about a future where "a few models eat everything they see." That phrase captures the bet companies are making right now whether they realise it or not.

View full thread

That's why our core technology is open sourced under MPL 2.0. Why on-premise hosting is easy. Why our services work like dumb pipes that move your encrypted data around on your behalf. We physically can't do anything else.

There is a twisted irony for companies right now. In Satya's latest article "A frontier without an ecosystem is not stable," he warned about a future where "a few models eat everything they see." That phrase captures the bet companies are making right now whether they realise it or not.

View full thread

Organizations need a place where agents and people come together around a piece of work. When your business processes can all speak the same language, the knowledge generated by one workflow becomes available to the next.

There is a twisted irony for companies right now. In Satya's latest article "A frontier without an ecosystem is not stable," he warned about a future where "a few models eat everything they see." That phrase captures the bet companies are making right now whether they realise it or not.

View full thread

This is what @tonk.xyz is designed for. A data sponge that absorbs all the decisions, data points, business signals, all that tacit knowledge, and makes it available to the model at point of use.

The loop becomes the new IP of the firm. Every workflow accumulates tacit knowledge unique to your business and compounds your token capital.

There is a twisted irony for companies right now. In Satya's latest article "A frontier without an ecosystem is not stable," he warned about a future where "a few models eat everything they see." That phrase captures the bet companies are making right now whether they realise it or not.

View full thread

It starts with model replaceability. That's the first golden rule of durable advantage. Treat the model like a contractor you can replace. Treat the knowledge layer like the company veteran you can't.

Once models become interchangeable, the source of differentiation shifts from the model itself to the knowledge and workflows surrounding it.

There is a twisted irony for companies right now. In Satya's latest article "A frontier without an ecosystem is not stable," he warned about a future where "a few models eat everything they see." That phrase captures the bet companies are making right now whether they realise it or not.

View full thread

The fix, according to Satya, is to invest in what he calls token capital. The AI capability your firm builds and owns. Accumulated knowledge, workflows, decisions, context that your AI systems learn by operating inside your business.

In order to achieve sovereignty, companies need an open ecosystem that can give them tools for success. You won't have a business unless you own the systems accumulating the knowledge that powers your AI.

There is a twisted irony for companies right now. In Satya's latest article "A frontier without an ecosystem is not stable," he warned about a future where "a few models eat everything they see." That phrase captures the bet companies are making right now whether they realise it or not.

Companies must rapidly adopt AI to survive. Yet that same adoption is setting them up for failure. The more they rely on external models to do the work, the more their advantages and margins disintegrate.

Like policy makers, working retroactively.

Boris@bmann.ca

I found a really great article around web standards "Every feature begins life as a non-standard proposal, often shipped. This is not a bug" that I shared with @blaine.bsky.social last September. Thread in the forum discourse.atprotocol.community/t/web-standa...

The industrial revolution organized companies around the production line. The computer revolution organized them around the database. What will be the organizing principle of the AI revolution?

View full thread

You can read it like a document, edit it like a spreadsheet and ask AI to reason over it like a system. Every interaction sends a signal back to the system; where people and machines cultivate a unified context through shared objects in order to achieve a common goal set by you alone.

The industrial revolution organized companies around the production line. The computer revolution organized them around the database. What will be the organizing principle of the AI revolution?

View full thread

This is what I'm working on at @tonk.xyz . Tonk offers a local-first unified view over your data and makes accessing that unified view ergonomic to both the human and the machine.

The industrial revolution organized companies around the production line. The computer revolution organized them around the database. What will be the organizing principle of the AI revolution?

View full thread

The tools we use today naturally separate context. They make working with context challenging. They naturally promote fragmentation, because this is how data was best organized for a computer. That’s historically been a blessing and a curse.

In order to unlock the full benefits of AI, we need new tools that allow us to bring together a unified view for a team of humans and machines, while keeping that context local-first and owned by the people creating it.

The industrial revolution organized companies around the production line. The computer revolution organized them around the database. What will be the organizing principle of the AI revolution?

Every era has a new bottleneck. In the industrial age, it was turning raw materials into goods. In the digital age, it was turning data into useful records. In the AI age, the bottleneck is context.