We'll be joining @RareEvo 2026 as a Featured Exhibitor next month in Las Vegas!
We'll be hitting it pretty hard, showing off what we've been cooking up the last few months. It will not disappoint!
Visit https://t.co/YX9KFHPK03 for a complimentary ticket
AI is very good at covering ground fast. Early on everything moves quickly and progress feels real. Then you get close to done and the curve just flattens out.
That last stretch is where the real problem shows up. AI can't evaluate when something is truly finished.
That's why it's important to have someone in the loop who can catch problems early and steer things before they go sideways.
In this video we walk through why that gap exists and what it takes to get across it.
Working with AI in practice looks very different from just prompting and accepting the answer.
AI can help you move faster, but it still won’t reliably tell you if something is truly correct or actually done.
Knowing whether that plan actually makes sense is where experience comes in.
Our CTO explains how this plays out in practice 👇
For those wondering how @midgardprotocol is progressing, I decided to try it myself.
Specifically, I tried from this commit, which @phil_uplc let me know was a decently stable snapshot, if you want to try it yourself.
https://t.co/2iFaEYN9aE
I was able to succesfully deploy some SundaeSwap contracts, make a deposit, create a constant product pool between two dummy tokens (VALHANILLA / BIFROSTING, for a bit of thematic punnery), and execute a swap of 31,415 VALHANILLA for 30,369 BIFROSTING.
The TxHash of the execution of that first swap is 68ae28279b0881ffd742b2208df6b2bf308101025a7a9eac45769442b34aec9d.
@phil_uplc confirmed that I'm the first person to run a smart contract more complicated than "always fails" or "always succeeds", meaning I can now claim the title of first to run meaningful contracts on Hydra, Leios, and Midgard. My trophy belt is getting quite full! :)
My honest assessment: progress is impressive, but developer experience right now is rough; there were quite a few things I had to hack through to get it working. The protocol is fiendishly complex, and so its success will entirely depend on how well the team can iterate and, as we say, "finish the second 80%" for a production grade protocol.
🛑🛑The vast majority of Midnight transactions could be completely FREE to end users.🤔
Projects can sponsor the DUST required to execute txs, meaning users may not even need to hold NIGHT.😎
Think entire applications and ecosystems where users don't even realize they're interacting with Midnight
Game changing. 🔥🚀
We’re excited to announce our newest feature: Credit Line 💳
LP assets can now be used as collateral on Strike 🔥
Take your LP tokens from the USDCx/USDM pair and deposit them into Strike to fund your account for trading or supplying liquidity.
This is the largest stablecoin pool on Cardano and people will be able to:
1. earn money from the dex liquidity pool
2. earn money from the Strike vault
Double yield, because everyone makes money on Strike.
More pairs coming soon.
Link: 🛥️ https://t.co/JnnDR0j57F
Full details: 🕹️ https://t.co/LRnbIaDf4l
We’re not just using AI to ship faster at Sundae Labs; we’re using it to build better, while protecting what matters most: ownership, quality, and team happiness.
Our CTO @Quantumplation shares the mindset:
- own every line of code
- use the time savings for deeper testing & stronger correctness
- raise the quality bar
- and turn extra capacity into real work-life balance.
Watch below 👇
When an LLM seems insightful or writes new code, it's mostly revealing subtle patterns it learned from training data, not reinventing truly novel ones. The plans AI generates may look structurally perfect, but for the untrained eye, subtle errors hide in plain sight.
That gap between plan and execution is crucial. This is why understanding how to use AI effectively matters, especially when writing mission-critical software.
The Gummiworm specification is now publicly available.
This document outlines the architecture, communication protocols, and wire formats that underpin the project, and serves as a foundation for peer review, technical validation, and future development.
Read the specification here: https://t.co/k3cVmmTdWa
Feedback and discussion are welcome.
Dingo development happens quickly. We're about 80% through our Q2 milestone deliverables.
As part of our testing, we created an example app using Dingo over UTxO RPC to swap tokens on the preview testnet SundaeSwap v3 contracts.
Start apps, connect a wallet, and trade.
AI is an incredible force multiplier, but in high-stakes environments like blockchain where we handle real $$$ on an immutable ledger, quality and caution are non-negotiable.
Join Pi this Wednesday for a special AI workflow livestream, showcasing how we at Sundae Labs have been leveraging AI to develop on Cardano.
May 27th @ 1pm EST
This coming Wednesday I'll be doing some live-coding.
Currently the plan is to talk about (and show off) how Sundae Labs is leveraging AI to accelerate development, without giving up the commitment to quality or craftsmanship that has made us successful.
I'm hoping to co-stream through the @pragma_org Discord, and the official @SundaeSwap Twitter account, I hope to see you there!
This coming Wednesday I'll be doing some live-coding.
Currently the plan is to talk about (and show off) how Sundae Labs is leveraging AI to accelerate development, without giving up the commitment to quality or craftsmanship that has made us successful.
I'm hoping to co-stream through the @pragma_org Discord, and the official @SundaeSwap Twitter account, I hope to see you there!
We’re doing a pop-up live coding session next week with Pi Lanningham of SundaeSwap Labs👀
Come hang out, watch real-time development, ask questions and see what the team is working on.
🗓️May 27
🕐1 PM EST
Join the Pragma Discord to follow along, link at https://t.co/yvY7UKbgIm
Smaller, cheaper Plutus scripts are now one command away. On recent mainnet scripts, the new uplc tool delivered 10%+ average cost savings.
Read more: https://t.co/O3Kz0PoQkV
The Plutus treasury proposal funds the next wave, read more here: https://t.co/JMyQp1CzQ6
The fix is merged. Bug closed.
But without being able to step through execution and see exactly what data the contract was reading, this would've taken much longer to find. Or worse, gone unnoticed.
That's the whole point of Gastronomy. 👉 https://t.co/k5yLqCUvoC
We used Gastronomy, our UPLC debugger, to find a bug in Aiken's tooling. One that was causing developers to see different behavior locally vs. on-chain.
Here's how we caught it. 🧵
That mismatch was the clue.
Stepping through execution one instruction at a time, we could see exactly what data the validator was reading and pinpoint the moment something wasn't right.
Fix: https://t.co/Ek4iNHtJLf