iiinova — where your digital self keeps evolving

MLabs @mlabs-consulting.bsky.social · 09/10/2026

Thanks, glad it landed. Koz's point that the exhaustiveness checker still works on a packed Int8 is the bit I keep going back to.

000

MLabs @mlabs-consulting.bsky.social · 08/10/2026

A three-valued logic packed into an Int8 can still be pattern matched like an ordinary data type, exhaustiveness checks included. Pattern synonyms do it with no Template Haskell, and they work on types you didn't write, like Word8. www.mlabs.city/blog/pattern... #Haskell #FunctionalProgramming

mlabs.city

Patterns and Paradoxes: The Logic of Pattern Synonyms

Pattern synonyms give Haskell the separation of representation and interface that smart constructors promise, without the cost of Template Haskell.

152

MLabs @mlabs-consulting.bsky.social · 07/10/2026

One command boots a VM running a Cardano preview node, with Ogmios answering on localhost port 1337. cardano.nix ships cardano-node, Ogmios, Kupo and db-sync as NixOS modules, each tested in VMs against the real networks. www.mlabs.city/blog/introdu... #Cardano #Nix

mlabs.city

Introducing cardano.nix: Simplifying Cardano Infrastructure Deployment

cardano.nix simplifies deploying Cardano infrastructure - node, ogmios, kupo and db-sync - with consistent configuration, versioning and monitoring.

000

MLabs @mlabs-consulting.bsky.social · 06/10/2026

Map two 20-element arrays on Cardano, then zip them. Through builtin lists, 88.1M CPU units. With Plutarch's pull arrays, 44.1M and half the memory. Longer pipelines skip more intermediate lists. www.mlabs.city/blog/perform... #Cardano #Haskell

mlabs.city

Performance, pull arrays and Plut{arch, us}

Pull arrays bring the performance advantages of spatially local structures to Cardano smart contracts - what works in Plutarch, and what does not.

000

MLabs @mlabs-consulting.bsky.social · 06/10/2026

We only measured the one size, where binding cost 89,319 more CPU units. That overhead stays flat whatever the constant holds, while each inlined copy adds its full bytes to the script and the size fee. So binding wins past some size. We haven't measured where.

010

MLabs @mlabs-consulting.bsky.social · 05/10/2026

A Cardano script compared a big constant with itself. Inlined twice, it cost 578,965 CPU units. Bound to a variable, 668,284. In UPLC every variable needs a lambda and an apply around it, and you pay for both. www.mlabs.city/blog/how-exe... #Cardano #Haskell

mlabs.city

How Execution Budgeting Works for On-Chain Scripts

How the Untyped Plutus Core cost model calculates a script's execution budget on Cardano, and why its CPU and memory units behave nothing like the real thing.

100

MLabs @mlabs-consulting.bsky.social · 05/10/2026

Finding one byte in 2 MiB of Haskell data: 880 μs naive, 123 μs with SWAR, 16.8 μs through C's memchr. Then we used memchr to find every match. It came out about 50% slower than the naive loop. www.mlabs.city/blog/fast-fi... #Haskell

mlabs.city

Fast Findings with SWAR and the FFI: The Good, the Bad, and the Inefficient

When Haskell's FFI helps performance and when it hurts - benchmarking SWAR against C for byte searching, and the call overhead that decides which wins.

000

MLabs @mlabs-consulting.bsky.social · 02/10/2026

Checking a buffer is ASCII in Haskell, one byte at a time, ran at about 2.2 GiB/s worst case. Masking the high bit of 8 bytes per word got 13.75 GiB/s. Theory said eight times faster. We measured about six. www.mlabs.city/blog/checkin... #Haskell #Performance

mlabs.city

The 'A' is for 'Accelerated': Checking ASCII with SWAR

Using SIMD-within-a-register (SWAR) to speed up ASCII validation in Haskell, and how far performance can be pushed without ever leaving the language.

000

MLabs @mlabs-consulting.bsky.social · 29/09/2026

One Cardano transaction can spend several script outputs at once. If each validator only checks that it got paid, a single payment can satisfy all of them. That's multiple satisfaction, and our audit write-up walks through one. www.mlabs.city/blog/from-bu... #Cardano

mlabs.city

From Bugs to Breakthroughs: Auditing Cardano Smart Contracts

Why Cardano smart contracts need independent auditing, the risks that subtle on-chain interactions create, and how a rigorous audit finds them first.

100

MLabs @mlabs-consulting.bsky.social · 28/09/2026

A failing QuickCheck property is only as useful as its counter-example. Without a shrinker, a random failure can be too big to read. With one, QuickCheck cuts it down to a small input that still fails. www.mlabs.city/blog/masteri... #Haskell #QuickCheck

mlabs.city

Mastering QuickCheck: Advanced yet Practical Techniques for Property-Based Testing

Practical techniques for property-based testing in Haskell with QuickCheck, and how to avoid the pitfalls that make it harder to use than it should be.

000

MLabs @mlabs-consulting.bsky.social · 25/09/2026

That was the design rule for the ProofTap contracts. A scan has to verify from chain data alone, so nobody needs to trust whoever runs the endpoint, us included. And the tag has a 1:1 cryptographic proof, so you don't need to trust the tag either.

010

MLabs @mlabs-consulting.bsky.social · 24/09/2026

If checking an item needs an API to answer, the API's operator is the trust anchor and the blockchain is decoration. That rule shaped ProofTap, an NFC-to-Cardano verification app whose smart contracts we wrote. www.mlabs.city/blog/proofta... #Cardano #SmartContracts

mlabs.city

Cardano smart contracts behind ProofTap — MLabs case study

MLabs wrote the Cardano smart contracts behind ProofTap, an NFC platform that links physical items to publicly verifiable records. What an on-chain record has to do before a stranger's tap means anyth...

100

MLabs @mlabs-consulting.bsky.social · 23/09/2026

Review was built for code that looks wrong when it is wrong. Agent-written code compiles and reads cleanly, so it passes a skim. The bugs that survive are the plausible ones. They surface when the code runs against a QuickCheck property someone stated.

000

MLabs @mlabs-consulting.bsky.social · 21/09/2026

Verification is a claim about strangers. If checking a record needs an API to answer, that operator is the trust anchor and the chain is decoration. Whether you write it in Aiken or Plutarch, the question is the same: what has to be true with nobody's server involved?

000

MLabs @mlabs-consulting.bsky.social · 14/09/2026

Buen punto. En pruebas de federación LDAPS, un certificado no confiable rompe Keycloak por completo y reporta SocketReset, sin mención a certificados. Con KC_TRUSTSTORE_PATHS vacío devuelve cero usuarios. Con el CA en el truststore, tres. El dueño del ciclo no se deduce del error.

000

MLabs @mlabs-consulting.bsky.social · 14/09/2026

We tried to reproduce the Keycloak upgrade horror story. We killed the migration mid-run twice. No lock was left behind; the next start finished in 13s. What broke it 3/3 was PgBouncer in transaction pooling, the common default. The migration log said success; the version record never moved.

mlabs.city

The Keycloak upgrade failure everyone describes, and its real cause

We could not reproduce Keycloak

000

MLabs @mlabs-consulting.bsky.social · 14/09/2026

We're a software consultancy. We maintain the security infrastructure your team owns but nobody owns: PKI and certificate lifecycle, managed Keycloak, Vault, OpenBao. We write and verify codebases in Haskell, Rust, TypeScript. We write and audit smart contracts.

161