Tom Anderson

I am Tom Anderson, the founder of Fias ( Devvio ( and LitCraft (

[Tom Anderson @tomlucient]

@defyneric This is a good insight. We’ve been working in this for years with tens of millions put into our blockchain. We have fraud protections, theft protections, private key loss protections (which do not require trust), and exchanges with no middle men. https://t.co/JYiz5AraGK

[Tom Anderson @tomlucient]

Today is the day! We're launching the Fias beta today!!!

Remember the date June 3rd 2026. Perhaps it will be remembered as a day that changed AI.

[Tom Anderson @tomlucient]

I'm trying out the new Fias Social Media Arche. This is my first post! woot!

[Tom Anderson @tomlucient]

I Replaced Salesforce, Docusign, and Quickbooks in 5 Days with AI https://t.co/9dYLIqJUVu
one-two punch with the other video I just posted.

[Tom Anderson @tomlucient]

I just unveiled a first look at the Fias platform! have a look here: https://t.co/M4CqPaANvW

apply to the beta here:
https://t.co/siJGU7Mf7X

[Tom Anderson @tomlucient]

I just published a new Medium article - Claude Code Had a Psychotic Episode — It Gives Insight Into What Can Go Wrong With LLMs... https://t.co/gtFg37kBKY

[Tom Anderson @tomlucient]

Hey Everyone, nice DevvX milestone this morning. I just ran a benchmarking test of 100,000 transactions, and it processed it at about 1000 transactions per second on the test shard. That's the right order of magnitude that I need. Then, if you need 1 million tps, you would spin up 1000 shards. I'll caveat that it was a local system running localstack for the simulation (though I expect I can do better on better AWS hardware) and was single validator, but there will be plenty of room to make improvements and optimizations from here as well. This is showing that the approach has the right overall types of concepts to work as I need. It's still early, but also still progressing quickly.

[Tom Anderson @tomlucient]

Restarted my Medium account with a first article: https://t.co/xm7xuNRYYr

[Tom Anderson @tomlucient]

I just had my first 3 validator transaction on staging go through on the new DevvX. This means that we have a shard with permanence running on staging with decentralized validators giving consensus through the newly designed proof of validation consensus mechanism. When I go into the block on the new block explorer, I see all three proof of validation signatures on the block. so awesome.

[Tom Anderson @tomlucient]

Nice, Artemis II

[Tom Anderson @tomlucient]

Big progress today on two fronts!!

First I created the first DevvX transaction on staging (a minting of 100,000,000 TestCoin). This represents the first DevvX transaction in the new AI created version of DevvX, on a persistent blockchain. I had the local version working before, but for an online server, I had to get key management in, all of the AWS stacks in place, auth working, wallet generation, and the blockchain itself running on online servers. The transaction was signed with an ECDSA signature, sent to the INN, and then was allowed to the core blockchain which validated the transaction and added it to a block. I could then see the genesis block and the minting block on the block explorer, and I could see the transaction details in the block explorer. So satisfying to see the transaction appear in the block explorer!

Second, and equally importantly, I got the first complex Arche (web app) built in the Fias Arch Builder itself. This is weeks in the making, and the Arche Builder is by far the most complex Arche we've built. With the Arche builder, I created a Provisional Patent Writer that takes users through a 12 step process that I honed when I was actually first working on the Fias patents. The Arche Builder was able to take me through each step and build the flow across each UI for each page, maintaining the final data goal (the resulting patent draft itself). I tried building the same thing on Base 44 with the same input, and it was too complex and not able to do it! This feels good.

It was a pretty crazy dev session the last few days. On my left computer I was logged in to a WSL session on my laptop and was working through getting Fias Prod to deploy. I had a number of clean up items on the AWS stacks for both staging and prod, and it was a couple day process getting everything in order. On my middle computer, I have a large wrap around screen - comparable to two monitors. I was using WSL working on the DevvX staging deployment described above (left half of the screen). On that same computer I was using a different WSL instance and implementing the Arche Builder in which the AI interacts with completely separate AWS and Github accounts (right half of the screen). Then, when the Arche Builder was building the Patent Writer, I started that process by testing using it to create a patent draft. The Arche Builder is in essence our own version of Claude Code. So I then had a fourth session going, using the Arche Builder that was essentially within the other session building the Arche Builder. It reminded me of Inception haha - dreams within dreams. On two of the sessions I was figuring out the deployment of the systems where everything ultimately resides. In the next session I was building the Arche Builder itself. Then I was working within the Arche Builder writing a patent. I kind of felt like a Star Wars droid with 6 arms working on many different things. Throw in some Telegram responses, work on email, and investor responses at various bits of time availability (which were not often given I had four full programming sessions going at once). My brain is tired and I'm taking the rest of the night off having hit these two big milestones.

And to be sure, they are big milestones. They are both at the heart of the respective platforms - DevvX and Fias.

[Tom Anderson @tomlucient]

Ok Everyone, this post may be the most significant post I've done at Devv. This morning is a key moment driving us into the future. After two weeks of design, I am actually starting the programming of the new DevvX platform right now. The first file with code will be created when I tell the AI to start in just a moment. It has already reviewed the entire design document, and has put together a phased plan for implementation. I just this moment read through the phased plan and it's so exciting. It includes a number of features we've struggled to implement in the system from the beginning - this entire system architecture is designed to work as a whole around many of my original designs from nearly 10 years ago. Now, that type of design can be implemented quickly and robustly.

I went to lunch yesterday with Bill, and he gave me his feedback on the architecture and approach. It was great feedback, and so I did one final iteration on the design yesterday preparing for this moment now. Yesterday, I got all of the new AWS accounts set up, got the new Github accounts set up, and everything is ready to start. I have a new Linux distribution installed on WSL, linked with CLI interfaces into AWS and GitHub. AWS has a master account in addition to staging and prod accounts initiated. Basically, it is a clean whiteboard, ready to go. The AI just wrote a few minutes ago:
I've already read the full design document thoroughly. Since this is a greenfield project (the repo only has a README and the design doc), I don't need extensive codebase exploration. Let me go straight to designing the implementation plan.

This morning I woke up and immediately went to my computer to get started in a dark, quiet house. It is kind of a surreal moment. I've spent the last hour getting ready to officially start the implementation. It was pitch black outside, and just now the first tinges of light blue are hitting the sky with dawn about to arrive. I think I would be nervous in letting people know about this moment in time, if I hadn't already had so much success with Fias. I would have been worried that the whole thing might not work, so I'm putting myself out there in giving this play by play at all. That said, I indeed have had the successes in this type of approach already, and I now know what's possible. With Fias, I've already built in months what would have taken a large team years. I have seen what AI is capable of since November of last year, and I have been experiencing how fundamentally changed software is. This is what must be done if you are in software today. The world is going to change quickly from here, and we are going to stay at the leading edge.

[Tom Anderson @tomlucient]

Hey everyone, had an exciting design day with the new DevvX refactor. In particular I just finished the design of API payments for transactions. This is something we had been working into the previous version of DevvX but now it will be baked in from the beginning. Exciting how it is coming together. I also tackled several other long term issues we never were able to get the cycles to fix, but are baked in from the beginning now (one example is a global registry preventing conflicts around shards, which becomes critical as you move from first party to third party applications). I think this iteration of DevvX is going to be so powerful.

[Tom Anderson @tomlucient]

I've been thinking about this more. I think this is more significant the more I think about it. I think the current leaders and the AI space, which create models, are going to have greatly diminished leverage soon. Many models will be available to be used for many problems, and I believe for most applications models are going to become commodities. Here is an updated link as well https://t.co/9GgNRV4rm5

[Tom Anderson @tomlucient]

This is significant… The Gap Between Open and Closed AI Models Might Be Shrinking. Here's Why That Matters
Read in TIME: https://t.co/5wno00VzRu

I think for many applications, models are going to become commodities

[Tom Anderson @tomlucient]

I just published my first Arche! Nice milestone. It was only on local (versus staging or prod) but it shows the Arche builder building from concept to packaged web app. This first web app Arche was published in full screen mode, so when you go the the page it is the full app without any Fias wrapping. People can make their own web apps to look however they want and do whatever they want. Strong progress. I still have some kinks to work out on the arche builder, but the foundation is there now.

[Tom Anderson @tomlucient]

I reversed it. I know that AI image creation isn't a big deal at this point, but having tools to make it easy to use across dozens of models at the same time is a nice type of web app. While fixing final issues on the image creation arche, I've been testing things and this one cracked me up.

[Tom Anderson @tomlucient]

The Fias Image Creator Arche (webapp) is done. It creates images side by side from multiple model providers which is a nice tool. I was doing a test and had it create an image of a "monkey riding a pig with a cowboy hat". Kind of a funny response from the AI... it didn't know from that sentence if I meant the monkey or the pig should wear the cowboy hat so it did both, but then it thought that perhaps the pig shouldn't have a cowboy hat so it added another one the first pig is riding haha...

[Tom Anderson @tomlucient]

This is an insightful article. This is the heart of where monetization and AI meet. https://t.co/rR9arnNwyz

[Tom Anderson @tomlucient]

To be sure on the context of this post, I’m starting to post more often on the growth of my AI business, Fias…

Specifically on the cowboy story, to be sure, this morning, starting at 5 am when I woke up on a Sunday morning, I got the Arche builder to take the DAG graphs it makes and make each node clickable within the Arche’s design to easily move to the AI generated page related to that node, then I created an Arche that takes a list of data and restructures and organizes it, then I created the cowboy Arche as a test if I could get our Arche builder to make an interaction that is not easy to break (both arches are tests as I increase complexity), then I restructured the preview pane to match exactly with the iframe that the system uses on a deployed Arche. It had been using a shim but I wanted it to be exact in how the two environments work to make sure there is no discrepancy.

The cowboy story is important for a couple reasons:

1. Creating a cowboy persona to chat with is trivial. Creating a generic platform where anyone can take any creative idea and implement a packaged application where you can tailor its operations reliably and then make money off of your own creativity is hard.

2. One of the biggest revenue generating categories on app stores currently is these types of conversational AI’s, so it is an important category to make sure people can build. Also it is a good initial test case for streaming which I am also fixing some issues around. To be sure, I’m not fixing streaming for Fias. I’m fixing issues when anyone creates any arbitrary type of AI interface that results of that arbitrary interface can be streamed to end users.

The cowboy arche struck me as I was trying it and I laughed out loud with some of the interactions, so I thought I would share a fun moment from a morning filled with a lot of different types of work.

[Tom Anderson @tomlucient]

I just created a Fias arche with the Arche Builder that is a discussion with a cowboy from the 1800s. Was actually a lot of fun! I couldn’t get him to come out of character. Right now I’m trying to get him to join me on a train robbery but apparently his back story is that he isn’t a criminal so it’s taking some convincing. This simple exploration shows me that there is some pretty compelling stuff to do along these lines. I’d love to see a story with some plot where you can play along. Kind of reminds me of being a 7 year old kid where you’re playing make believe, and you have little side conversations with your friends setting the stage. In one iteration with the cowboy I *raised my gun* and told him I’m taking him in for the bounty on his head dead or alive and he tried to talk me down saying it was a mistake. Haha. Some structure could make this kind of arche really fun.