Week 42 · 2026

Highlights

cardano-init v0.3.0 shipped (D3, #92, #93, #95)

v0.3.0 was published on Monday, with binaries for Linux, macOS and Windows and an npm installation path. It packages the docs site, doctor improvements and shorter flags described last week, together with two newly merged features from Paulo Bressan:

• #92: a default spec/ folder. Generated projects get a place to describe actors, assets, transactions and invariants before changing the code. Projects with on-chain, off-chain or fullstack components also get a worked gift-card specification. --no-spec skips it, and later add/remove operations leave the authored specification alone.
• #93: an update notice before generation. When a newer release exists, the CLI suggests the update command for the installation method. The lookup is cached daily, fails quietly and is disabled for JSON, CI and non-terminal output.

The release-preparation PR, #95, records 220 passing tests alongside formatting, linting and distribution checks.

Separately, @lisicky opened #97 to configure MCP servers in generated projects. The proposal adds .mcp.json and agent instructions for Cardano Debug transaction validation and script debugging, plus an opt-in Yaci DevKit endpoint for devnet operations. A registry controls which servers are offered for each stack, and add/remove preserve servers the developer configured by hand. The PR remained unmerged; the Yaci endpoint requires its 0.12 beta.

Hardfork readiness gained a dependency-evidence prototype (D1, D8)

48 of the tracker's 54 commits are on the unmerged hardfork-readiness branch. The work asks a practical question: when an upstream component changes, which Cardano tools and developer workflows might depend on it?

The prototype creates coverage records for all 269 tools in the ecosystem map and collects dependency evidence against pinned source revisions and explicit build configurations. Native collectors expanded across JavaScript, Rust, Go, Python, Haskell and Java tooling. The collected profiles reach beyond libraries into wallets, governance applications, signing tools and chain-data services, including Lace mobile, GovTool, Cardano JS SDK and Cardano GraphQL.

The distinction between a declared dependency and a deployed application matters here. A wallet's JavaScript lockfile does not establish which native binaries, hardware firmware or hosted services it uses. Those boundaries stay visible as gaps. Dependency paths retain their consumer and configuration, so a version used by one project is not silently substituted for another's.

Static coverage and impact pages, downloadable evidence, snapshot comparisons and a resumable refresh pipeline are implemented on the branch. Review packets prepare the next step: tracing actual capabilities and service integrations through the retained source.

Every tool remains hardfork-unassessed. Dijkstra was selected as the target, but the exact node source pin and test environment were still unresolved at the end of the week. This is groundwork for compatibility assessment, with target-specific testing still ahead.

A maintainer survey and a proposed scoring method (D1, #176)

Hinson Wong's draft #176 adds Delta Lab's survey instrument and evaluation methodology for the tooling-evidence workstream.

The survey asks one response per tooling project about dependencies, downstream users, onboarding friction, AI-assistant support and willingness to review contributions. It complements the Cardano Foundation's individual developer survey. Node 11.2 compatibility and expected readiness are separate questions, with follow-up on what a project needs to become ready. Project attribution and quotations each require their own permission.

The proposed score weights survey usage at 60%, GitHub stars at 10%, and map connectivity at 30%, with rankings confined to each category. Missing usage produces a score range rather than a zero or a rank: only 73 of the 269 projects have a usage figure in the chosen survey snapshot. The draft also records how conflicts involving Delta Lab's own tools would be handled. These are proposed instruments and rules; no new survey results were delivered this week.

Smart wallet restrictions became usable examples (D5, #55)

Sofia Bobbiesi's 14 commits expanded the smart wallet's delegated authorization scripts and opened #55:

• Recipient allow and deny lists check either the full address or payment and staking credentials separately.
• Asset allow and deny lists constrain which token quantities may change during a spend, with corresponding deposit scripts controlling what enters the wallet. The comparison catches partial movements as well as whole-token removals.
• A spending-period limit caps cumulative ADA outflow over fixed periods, such as a day or a week. The datum carries the period boundary and amount spent, and the script checks both the cap and the next state. Fixed boundaries keep the period from drifting when a wallet is idle.

The changes include Aiken tests, MeshJS constructors and encoders, devnet cases and usage examples. A final integration fix accounts for the provisional change address used during transaction fee estimation, which strict recipient lists must accommodate while the builder evaluates the transaction. The wallet PR was still open at the reporting cutoff.

The tokenized bond gained issuer-triggered claims and multiple vaults (D5, #54)

Francisco Joray's nine commits completed more of the redesigned bond's on-chain and off-chain paths in #54.

• Issuer triggering is opt-in. A certificate owner can commit a payment credential, allowing the issuer to trigger coupon claims and maturity on their behalf. Payments stay bound to that credential. The owner can revoke the permission, and transferring ownership clears it.
• One instrument can use multiple vaults with different terms. Each certificate records its selected coupon policy. Vaults are authenticated by an NFT, and deposits and redemptions must preserve that identity in the continuing vault output.
• The off-chain implementations caught up. MeshJS builders and the Tx3 protocol were updated, with usage documentation. Mesh devnet scenarios now exercise owner and issuer paths, multiple vaults, and rejected transactions such as premature claims, overpayment and issuer actions without consent.

Both contract branches also pinned Yaci DevKit 0.10.6 for reproducible devnet runs after the newer launcher disrupted the test setup. The bond PR remained unmerged. Vault funding and custody remain explicit assumptions; the new checks do not establish future solvency.

The property catalogue and vesting review opened (D5, #52, #53, #56)

Dr. Mazen Khaddaj opened two draft milestone deliveries from No Witness Labs. #52 establishes the property catalogue: shared schemas, validator snapshot metadata, authoring guidance and six sample Universal entries referring to twelve Lean theorems. Separate checks validate the JSON, compile the selected proof module, resolve theorem names and inspect axiom dependencies. The examples concern the imported evaluator and V3 ledger models; their scope and SMT trust assumptions are recorded explicitly.

#56 starts the linear-vesting review by pinning the upstream source, compiled script and compiler metadata. Its delivery currently consists of that snapshot and the review scope. Proof coverage, unfinished properties, reproducible examples and IO acceptance remain on the checklist.

#53 proposed fixture-generation fixes and batch proofs aligned with separate vesting continuations. It was closed as superseded, with its commits carried into the contributor's replacement branch. The closing comment explicitly leaves issue #51 open. Closing that duplicate PR did not deliver the formal-build fix upstream.

Advanced onboarding reached assurance and production (D4, #2059)

The Advanced branch gained drafts of its remaining two lectures in #2059:

• Higher assurance moves from one test per imagined attack to properties checked against generated transactions. The splitter and vault exercises show how a plausible fix can pass individual tests yet fail a broader property. The lecture then explains what audits and formal proofs add, and where their guarantees stop.
• Going to production takes the oracle example to a Preprod deployment workflow, with operator-key setup, deploy/update/monitor scripts, deployment records and a browser application. It covers providers, operational failures and the need to plan migrations when a contract changes.

The optimization lecture also gained a protocol-level perspective: choosing how state is distributed across UTxOs affects contention and transaction size before any local code optimization begins. All of this remained on the unmerged onboarding branches.

Design patterns grew to nine runnable projects (D4, #2063, #2062)

The Design patterns PR expanded from last week's three patterns to nine. Parameter validation, merkelized validators, linked lists, Merkle trees, Merkle Patricia tries and validity-range normalization joined the transaction-level and indexing patterns.

Each project pairs Aiken validators with Mesh transaction builders, attack tests and measured costs. The pages compare the straightforward implementation with the pattern and explain the trade-offs. Examples include recognizing parameterized scripts by reconstructing their hashes, splitting registries across authenticated UTxOs, and keeping a large registry off-chain while checking proofs against its on-chain root. Committed costs.json files feed the displayed measurements.

#2062 replaced its fixed CI project lists with automatic discovery of every aiken.toml and package.json under the onboarding examples. New examples therefore enter the test, build and blueprint-check matrices automatically. The main onboarding and design-patterns PRs were marked ready for review on Friday; neither had merged into the published portal by Sunday.

Two signatures and earlier deliveries recorded (D1)

Sebastian Nagel, Technical Architect at Input Output Global, and Andreas Osberghaus, Software Engineer at utxo AG, joined the pledge through #173 and #175. That brings the merged list to 38 signatories. Andreas also reported that the signing button had opened an issue, then supplied the PR adding his row.

The tracker recorded DX.02 delivered on August 24 and DX.03 delivered on September 23, and backfilled earlier Reactive contributions. These are updates to the delivery record, not new October deliveries.

A note on the counters

The five merges are three cardano-init PRs and two pledge signatures. The seven opened PRs are the survey, MCP support and five contract-library proposals, including the subsequently superseded #53.

The 93 commits split into 54 tracker, 23 contracts-library, 15 Developer Portal and one cardano-init. Of the tracker commits, 48 belong to the unmerged readiness branch. The two CLI features merged from last week's commits, so their merges do not add new non-merge commits this week. Contributions living only on forks, including the survey, MCP and formal review work, sit outside the gatherer's repository-branch commit counts.

The zero release counter also needs context: v0.3.0 was published by github-actions[bot], which the gatherer filters out. The release itself is linked above.