2024
We are living in an interesting polarity between a digital renaissance driven by AI and real-world set of shifts; changes with regards to geopolitical relations, wars, political division, technological challengers to the status quo. We are seeing the emergence of powerful AI for multiple modalities; text, audio, music, voice, images, video all at once. We are building new financial and governance systems that are open, decentralized and consensus driven; and the resistance from incumbents threatened by these systems. We are seeing the shifts from bits to bytes and the commodization / renaissance of digital creative universe, versus the bits that can work along side these software systems. We are seeing that the community driven, discord chats with remote teams are building and using systems along side the centralized closed in office teams building systems. It is an interesting time to be building and exploring the future of technology and the implications it will have on the real world. I am writing software with synchronous and asynchronous AIs in the editor and the browser. I am communicating with my personal AI on my local machine with the WiFi off.
This past year:
- I wrote a lot of react, node.js serverless APIs and AI agents
- I learned more about React Server Components and how to build with Server Actions
- I learned about RAG (Retrieval Augmented Generation) and Vector DBs
- I learned more about Zero Knowledge Proofs, how they work at a low level and in practice
- I learned more about MEV (Maximimum Extractable Value) and SUAVE
- I learned more about fine-tuning GPTs
- I learned more about nanoGPT and how to fine-tune and optimize a modal
- I learned more about open-source LLMs and how to run them on local machines
- I learned more about musicgen, MBD and how to generate music leveraging models
- I learned how to build multi-modal AI applications
- I learned how to build smart agents
- I learned more about how to build and sell software
- I learned more about the Compound Startup
- I learned more about PLG (Product Led Growth)
- I learned how to speak better Mandarin
- I learned more about the Rust toolchains in the Ethereum ecosystem
- I learned more about shared sequencers, L2 blockchains and distributed transaction ordering
2023 Year in Review:
This past year I went to AI hackathons focused on musicgen, elevenlabs, multimodal hackathons and using AI to build applications. I attended the Stanford Blockchain Conference and learned about distributed consensus prototols, shared sequencers, I attended Google Next and learned more about applied generative AI and NVIDIA frameworks for training and serving compute for inference. I spent a lot of time writing RAG AI applications for customers to automate post-purchase customer service workflows. I learned a lot about selling software and building products, listening to customers and building a pipeline. I spent time learning new parts of the ethereum developer tool kit including Reth, Foundry, Alloy, Rivet and more. I spent time rebuilding some of my eth transaction APIs from node to rust. I spent alot of time building frameworks for updating and retrieving data from vector databases. I optimized many parts of my software platform leveraging the latest with React Server Components (RSC), Server Actions and more. I spent time learning about how to create RAG based applications with local agents, and a good amount of time on prompt engineering and how to create different abstractions to have the model give the optimal output.
Here are the past couple years of this same blog posts.
2013: A Year in Review
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
This year I read and reread some great books:
Deep Work - Calvin Newport
Chip War
Becoming Supernatural -
The Price of Time -
Quantum - Miko
Size - Vaclav Smil
How the World Real Works - Vaclav Smil
Life Ascending
Play Nice But Win - Dell
Here was my focus for 2023:
- Learn & Gain Knowledge related to crypto programming in rust
- Learn & Gain Knowledge related to scaling software companies
- Learn & Gain Knowledge related to zero-knowledge proofs
- Learn & Gain Knowledge related to dag based consensus protocols
- Learn & Gain Knowledge related to numerology and astrology
- Learn & Gain Knowledge related to machine learning and generative ai
- Learn & Gain Knowledge related to building distributed systems
- Leverage IBC for connecting the Stateset Blockchain with other networks
- Spiritual Growth related to Meditation and Manifestation
- Read more whitepapers and research
- Write more on substack
- Learn and Speak Mandarin everyday for a 30 minutes a day
- Run, Box and Lift everyday
- Fast from 8pm to 2pm everyday
- Take NMN and Resveratrol daily
- Cold Showers
- Keep a daily journal and a physical calendar
- Say my autosuggestion every morning and every night
- Eat Healthy (Eggs, Leafy Greens, Mushrooms, Quinoa, Black Beans, Avocado, Turkey, Chicken, Elk, Steak, Fish)
- Drink more water
build, keep, learn, write, read.
Some highlights from 2023:
Tuscany with Family
Scotland and London for Anniversary
Hiking to Glacier Point in Yosemite
Drift Fishing in the Snake Rive
Joined the Club
Focus for 2024:
- Write great software and build great software products
- Explore more about generative interfaces for software
- Learn more about AI and how to apply bytes to bits
- Learn more about open-source models and build more smart agents
- Spiritual Growth, Meditation and Manifestation
- Learn more about Numerology and Metaphysics
- ReRead Autobiography of a Yogi
- Learn and Speak Mandarin everyday for a 30 minutes a day
- Run, Box and Lift everyday
- Fast from 8pm to 2pm everyday
- Take NMN and Resveratrol daily
- Cold Showers
- Keep a daily journal and a physical calendar
- Autosuggestion every morning and every night
- Eat Healthy (Eggs, Leafy Greens, Mushrooms, Quinoa, Black Beans, Avocado, Turkey, Chicken, Elk, Steak, Fish)
- Drink more water
build, keep, learn, write, read.
On Stateset
This year we spent a lot of our time redesigning and rebuilding our 3 core software offerings. We are continuously learning from our customers and are moving the company toward what resonates well with customers. We are not building in a vaccum but are closely listening to the market.
ReSponse CX
In this thought, we focused on taking ReSponse off the shelf from when we originally built it as a Shopify app in 2021 on GPT3 that can be used to generate responses to tickets that can be edited and then sent back to a ticket manually. We redesigned our standalone platform to be a complete offering that allow for building and managing vector embeddings and storing those embeddings based on organizational namespaces. We are also built in the ability to have rules engine that abstracts away custom system prompts. We focused on building multi-modal capabilities into the platform and making it as easy as possible for customer to self-service onboard into the platform. We are working with some great new customers that are leveraging ReSponse autonomously to respond to tickets, pulling in real-time data from systems like Shopify and also making function calls for post-purchase workflows. ReSponse has given us a great path to net new customer engagements and faster product development feedback cycles.
StateSet One
StateSet One is a core platform offering for DTC brands that need an OS for running Operations. We redesigned and rebuilt our core StateSet One offering on the latest React Server Components framework and also launched a single version of the platform also based on organizational namespaces. This shift towards a single version of the platform is more scalable and much easier to deploy feature updates across all instances. We are continuing to optimize the platform for multiple users including Customer Service team members and Warehouse Operators. We have optimized our search, our analytics and our APIs for interacting with our core platform.
StateSet Cloud
In doing so, we needed a better way to manage our infrastructure and give agencys / developer partners a way to integrate and build on our APIs. StateSet Cloud redesign has been focused on an IDE with deployment and log capabilities akin to how we can manage these on Vercel and Heroku. The move towards Kubernetes Engine has been an improvement on our devops automation and our ability to quickly deploy net new services on demand for customers. We are going to continue to offer additional compute offerings including GPU workloads for fine-tuning and inference.
StateSet Network
The perfect stateset...
what is the desired, optimal system a for globally replicated geographical diverse state machine?
what traits does the system have?
- verifiable finality
- geographic diverse set of nodes
- encrypted transaction mempools
- turing complete programability (serverless functions like programming paradigm)
- support for multiple programming languages to write functions
- fast confirmation times
- can sync state quickly
- can run it on a personal computer (without taking up too much disk space / too much CPU resources)
- rewards for running a node in the network
- slashing (penalties for trying to cheat the system)
- can join, leave and rejoin the set of nodes at will
- an even initial distribution of rewards
- the system is upgradable
- the system changes can be voted on (governance / voting mechanisms)
- the system is open-source
- markets can be quickly created
- support for multimedia (ability to store text, images, videos and songs)
- state can be quickly and easily verified (etherscan like block explorer like dashboard)
- fast recoverability of the system (TCB like coordination / notifications)
- graphQL query language on top of the system (the graph like subgraphs)
- access to real-world data that is offchain (chainlink like oracles)
- formally verified security of the system
- interoperablity with other blockchains and systems of record
- known and celebrated genesis time
- native stable and fungible unit of account
what are the user desires of the system?
- easily can be understood how to use (abstractions / concepts are sound)
- can run multiparty applications
- can run zero knowledge applications
- can self-custody rewards
- can self-recover keys to rewards
- optionality around mobile, web and desktop clients
- anonymity, psuedoanonimity
- reputation system
- insurance system for transactions (confidence in recourse if the system breaks)
what are the community desires of the system?
- meetups in every city every evening
- in-person hackathons every weekend (competitions for best apps developed)
- online community and wiki
- guides and tutorials to help new users onboard and build
- low learning curve to grok the system and how to build on the system
- support in multiple speaking languages
- new found land feeling (not too crowded)
- friendly and welcoming members
- benefits, advice and learning from the best
Thoughts on Building, Keep it Simple
1. being able to back prop
2. look at the world
3. try something
4. figure out how you are wrong and how wrong you are
5. and then nudge your company in the right direction based on how wrong you are.
Put it to 100 people and see what resonates.
We think people are going to resonate towards x, they resonate with y, shift the company towards y.
if you do that enough, quickly enough.
can get to a solution that finds product / market fit.
either we are moving to slow where we don’t have the feedback loop.
On MEV
I learned a lot about MEV at the Stanford MEV workshop and how blockchain protocols are a core infrastructure level have certain inherent advantages for infrastructure providers that have control over ordering transactions. This phenomenom is not local to blockchain systems but in AdTech, Ticketing Systems, Trading systems in more. Transaction ordering / censorships can be gamed for financial gains. The latest advancement on this has been Suave. Suave refers to a complex and innovative approach aimed at addressing the challenges of Miner Extractable Value (MEV) in blockchain systems. Suave, standing for Single Unifying Action for Value Expression, focuses on creating a more decentralized, fair, and privacy-preserving environment for handling MEV using privacy preserving TEE powered by Intel SGX. SUAVE offers a L2 platform and set of tools for building decentralized, privacy-preserving applications that can operate efficiently in the MEV landscape, aiming to bring fairness and balance to the process of block creation and transaction ordering in blockchain networks.
On ZKPs
I took part in the ZKP MOOC this year and learning more about how to build Zero Knowledge applications.
ZKPs a way to convince a verifier that some computation is true.
Proof is Short & Fast to Verify... Not only is the proof fast to verify, I can convince you without revealing the data.
Verifier knows simulated views and real interactions are indistiguishable.
What does it take to develop a zkSNARK application?
The steps include choosing a problem, generating cryptographic parameters and keys, constructing a proof for the problem, and verifying the proof.
ZkSNARKS understand algebraic constraint system. Write, Express, Code : your computation in a high level manor using ZoKrates, Noir, Circom, Gnark, Cairo etc. and that gets compiled down to the constraint system needed, then you need to decide which zkSNARK do you want to use (different SNARKs to use listed below). Lastly, need to choose algebraic object and find and optimized implementation.
The axis are:
Prover time
Proof size
Verifier time | yes or no
Trusted Setup preprocessor
circom - clear constraints, elegant syntax, limited abstration
arkworks - clear constraints, expressive as rust, manual opts
ZoKrates - easiest to learn, elegant syntax, limited witness computation
On Smart Agents
We can have a personal AI on our computer talk to us with out needing to be connected to the internet. It's amazing how we can interact with these technologies from a laptop. We can interact with an open decentralized global state machine leveraging a personal intelligent agent that is on our local machine. There is this shift the machines we have can are now powerful enough to be used to run these sorts of software applications. I am continuing to build on my knowledge of cryptocurrency and applying it to building intelligent smart agents.
On Bitcoin
Bitcoin is the 0, 0 of the money matrix. The idea of Bitcoin. Monetary Premium Chain, Secure, Censorship Resistance. Bitcoin extends one's time reference by creating a confidence in the store of value and therefore calcified incorruptibility of one’s time and energy savings. In doing so, trust in digital property rights becomes a means of increasing the predictability of one’s future. Bitcoin has another 10x-100x in front of it. Very hard to find 10x at $1B. That act of transferring a bitcoin is that which creates the value. Bitcoin is simply a namespace for the act of quantum transactional state verification. It is a wave until that act of spending makes it material. It is untangible yet verdical. Bitcoin is it.
Definiteness of purpose, the knowledge of what one wants, and a burning desire to possess it. I will continue to build the mastermind group and devote my energy to drive combinatorial creativity through technology and people.
Happy New Year from the Caribbean.
Live the Dream.
-Dom