We’re excited to announce Ora: the search engine for blockchain data ▩ ▩ ▩
Our vision is to make querying blockchains as easy as a Google search 🔍
Traversing crypto transactions should feel like surfing the web(3) 🏄
What’s the problem we’re solving ❓
Right now, it's impossible for most users to understand what's happening on-chain
Simple questions like "how much did I pay @MagicEden in fees last week?" require complex solutions (e.g. indexing an archive node or running SQL queries) 😰
Lots of people like @andy8052, @zachxbt or @jackbutcher have noted that Solana transactions are notoriously difficult to interpret
Why are all these accounts here, what's actually happening in the transaction, why is it calling into this program?
In two ways:
1️⃣ A blockchain search engine 🤖
2️⃣ A human-readable transaction viewer 🔍
We let users query on-chain data with natural language 💬
Using large language models in the backend, we let anyone easily explore, navigate and inspect on-chain transactions 🧐
We also built a transaction viewer your parents could understand
Want to see all trades between two dates? One click ✅
Want to filter for transactions that called a specific instruction? One click ✅
Want to order transactions by descending SOL amount order? One click ✅
Our first prototype looked like this lol
But we quickly iterated and soon arrived at a working conversion process for basic queries that hooks up into our search interface
As @FurqanR likes to say, we made sure to “get the f** out of localhost”
After a month of building for Solana's Summercamp hackathon, we now have a search engine that lets users
◼️ make complex queries in natural language - e.g. “show me all successful @JupiterExchange swaps between between 42 and 420 SOL from two days ago”
◼️ filter transactions - e.g. by called instructions or time range
◼️ sort transactions - e.g. by descending SOL balance changes
◼️ see summary descriptions in the form of “hey, X user called Y program with Z arguments, ABC is what happened”
So, what about existing solutions?
Block explorers and SQL dashboards are great ✨
But they mainly serve technical users and are not geared towards mainstream users 🤷♂️
Ultimately, Ora is a complementary piece to existing infrastructure 🧩
Our search results could further link to existing Dune/Nansen pages (@ASvanevik@mewwts if you’re reading this, we’d love to chat!) 🌐
Why now?
🚀 Web3 is slowly reaching a mainstream market
🔍 Blockchains are a radical shift in data transparency
🤖 Large Language Models are turning into APIs
📲 Crypto is going Mobile
While crypto hasn’t found its iPhone moment yet, initiatives like @solanamobile are leading the charge
Our hypothesis is that mobile users will seek out a Google-like experience in the long run
Or would you want to use an explorer / run SQL queries on a phone? 🤔
The idea of a web3 Google/search engine is not new
Neither is the concept of using LLMs for search engines
People like @amasad, @KyleSamani, @0xdoug, @ericjang11 have all pondered how LLMs could be leveraged in new search engines while @paulg has contemplated the value proposition of verticalized search engines
.@balajis has even floated the idea of leveraging LLMs to power a search engine for blockchains 👀
Coincidentally, that’s exactly what Ora is:
We let users query on-chain data with natural language.
Excited to share that we're launching dune.ora.so today!
A new search experience for @DuneAnalytics powered by LLMs.
Now you can find dashboards by describing what you're looking for - like "how's dex activity on eth?" - rather than relying on keyword matches.
Dune wizards 🧙 have created many amazing dashboards but they're not always easy to find today.
Queries like "magiceden opensea" don't return any results on Dune because the search requires 1:1 keyword matching and "magiceden" is interpreted differently from "magic eden".
Ora's approach solves this by creating a unique embedding for each dashboard
We run each dashboard through an LLM to contextualize it and then rank dashboards relative to a user's query based on semantic similarity
Large Language Models could drastically change how users interact with crypto applications.
From conversational protocols to smart social search, the whitespace for exploration is infinite. A few thoughts on ideas and risks 👇
1️⃣ Using AI to analyze blockchains
In the future, everyone will be able to ask any on-chain related data questions and get an answer.
Just ask in plain text and the AI model converts the question into SQL, runs that code on indexed data and surfaces the result back to you.
2️⃣ Talking to Protocols
Users will be able to transact on-chain by chatting with protocols (h/t @gaby_goldberg for pic).
Analogous to Siri, you'll be able to ask Ora to {bridge USDC, make an offer on y00ts with Nouns glasses, write a Mirror post on AI x web3 in your voice, etc}
Amidst the drama around Celsius, 3AC, Solend and a massive market drawdown, one of the strangest Solana MEV instances went fairly unnoticed
One token swap led to a cascade of arbs & liquidations that netted MEV bots more than $2 Million...
All within a span of 40 seconds 🧵
There are 4 protagonists in this story:
1️⃣ 7p..UU: the scnSOL whale (lost $700k)
2️⃣ G6..AV: the SOL-scnSOL arb bot (profited $25k, but lost out on $530k)
3️⃣ 44..Qu: the USDC-scnSOL arb bot (profited $400k, but lost out on $100k)
4️⃣ Fx..mY: the scnSOL liquidator (profited $2M)
What is going on with Solana, why is congestion an issue and how are the current TPS problems being addressed?
Strap in and get ready for a looong thread 🧵
As a quick recap, Solana has been fighting congestion issues for the past few months, experiencing two network halts (one in September 2021 and one in May 2022) as well as undergoing numerous periods, in which transactions regularly fail.
Taking a step back, Solana doesn’t have a mempool.
Say what?
Yup. Unlike our good friends in Ethereum-land, Solana uses a mempool-less transaction-forwarding protocol called “Gulf Stream” (maybe time for a rebrand? @aeyakovenko).