“Unlimited scaling,” “governance first,” “an academic chain” — around Cardano, the two phrases I’ve heard longest are the ones calling it a braggart and a vaporware artist. But this year, after checking the official roadmap and progress against each other one by one, I have to own an uncomfortable fact: the bragging is being delivered on, one item at a time. And along the way there was one genuine design change — and that single event says more about how Cardano ships than any slogan.

This article gets two things straight: where the chain is headed, and why anyone should believe it gets there.

First, the roadmap: three steps, none stalled

Cardano’s upgrade spine runs through a clear backbone. In September 2024, the Chang hard fork activated the Conway ledger era and put CIP-1694 on-chain governance live: changing protocol parameters or kicking off a hard fork became an on-chain voting matter, no longer a team’s call. In July 2026, the van Rossem hard fork pushed mainnet to Protocol Version 11 — the first upgrade approved entirely by on-chain governance, with DReps, stake pools, and the Constitutional Committee each voting; it updated the Plutus cost model (cheaper contracts) and paved the way toward Leios. From late 2026, the Dijkstra era begins in two phases: phase one ships Linear Leios (throughput) and Nested Transactions (resource savings); phase two ships Ouroboros Peras (fast settlement).

flowchart TB
    A["2024.09 Chang hard fork<br/>on-chain governance live"] --> B["2026.07 van Rossem<br/>first fully voted upgrade"]
    B --> C["late 2026 Dijkstra phase one<br/>Linear Leios + Nested Transactions"]
    C --> D["2027 Dijkstra phase two<br/>Ouroboros Peras"]

Note the sequence: governance first, then scaling. It’s not random — hand ownership to the community before operating on the body. van Rossem proved the path works; Dijkstra will walk it the same way. Those who laugh at it being “slow” miss the point: it refuses to be fast without a vote.

A three-way fork: it took the least mainstream road

Ask “how does a public chain grow,” and the industry already splits three ways, differing in which layer gets strengthened. Ethereum moves execution to L2 and leaves L1 as settlement and ordering — Vitalik’s “rollup-centric roadmap,” built on the belief that L1 can’t scale, so hand the work down. Solana goes the opposite way, pushing single-node performance to the limit and trading hardware for throughput — one behemoth machine, problem solved. Cardano takes the third road: rewriting the protocol structure, with Leios splitting ordering from packing and Peras adding a voting layer to consensus, betting that L1 can get fast by design, not by compromise or firepower.

My take: Ethereum’s answer is moving work to L2, Solana’s is stacking hardware, Cardano’s is redesigning the protocol itself. The third is the rarest and slowest to pay off — it needs papers, CIP reviews, and votes, none skippable. It’s betting on making L1 itself stronger rather than moving the problem elsewhere.

Delivery record: what was said, landing

“Academic chain” used to be an insult in my book. Not anymore. Line up the timeline and the papers interlock with shipping: the promise of provably-secure PoS produced the Ouroboros series, and today it runs Praos; the promise of on-chain governance landed as the Chang hard fork, governance genuinely on-chain; the promise that governed upgrades need votes made van Rossem the first fully vote-approved upgrade; the promise of a throughput breakthrough brought June 2026’s public Leios testnet (Musashi Dojo) and an August first run peaking at about 6x mainnet.

These span three-plus years, each step clearing a governance gate. So “academic chain” isn’t an excuse for not shipping — it’s the rhythm for shipping: argue first, then write the CIP, then test, then vote, then mainnet.

Closest to the blueprint in a decade

At this point I want to offer a more personal observation: today’s Cardano is closer to its original design blueprint than at any moment since it was founded.

I used to have a fixed impression that Cardano’s “roadmap” was repeatedly revised — redrawn ever further away. But in recent years, going back through its early design documents, I found the opposite: it never left the first drawing; it simply finally reached the positions drawn on it. The blueprint insisted from the start on provably-secure, formally verified — the later Ouroboros papers and Peras’s Agda spec are extensions of that exact line, makeup lessons rather than inventions. The blueprint’s layering idea (settlement separate from computation) now shows up in Leios’s RB/EB structure: ordering decoupled from packing is the most concrete fulfillment of the word “layering.” The blueprint emphasized community governance, no single point of authority — and from CIP-1694 to van Rossem, mainnet upgrades now pass through on-chain votes.

In other words, many chains’ problem is “the blueprint got changed beyond recognition”; Cardano’s problem is reversed — the blueprint stayed untouched for a decade, and only today has the progress caught up to it. Cardano used to exist in the papers; now it runs as the design vision. That’s also why I dare to make a judgment here: this isn’t improvisation, it’s a decade-old route finally reaching its payoff point.

On delivery, Leios actually redesigned once

To stay honest: this was no straight push. Leios is where a real design change happened.

The original research-paper variant was a three-layer structure: Input Blocks (IBs) carried most of the traffic; Endorser Blocks (EBs) endorsed the IBs; Ranking Blocks (RBs) ordered on Praos. Strong decoupling — diffusion fully independent of block rhythm — at the cost of a 5Δhdr window for equivocation detection.

At engineering, after absorbing a flood of community and CIP-editor feedback, CIP-0164 did the subtraction: fuse IBs into EBs, going two-layer — EBs both carry and endorse, RBs keep ordering. That’s today’s Linear Leios. The window dropped to 3Δhdr, security stays equivalent, and the structure got far simpler.

The before-and-after in one line: the paper chased maximum decoupling; the engineering version chased shipping first. In Linear, an EB must be certified and chained before the next RB arrives — decoupling is only “partial” — but as long as the network is healthy, an EB can be much larger than an RB, and throughput still rises. It also genuinely tripped during engineering: the September 2025 monthly review lists the handover from research to engineering, constructive editor and community feedback on CIP-0164, clarifications and cost analyses, plus a simulator bug found and fixed.

Will it change again? Yes, and that’s not embarrassing

I gave this real thought, and I’m confident: more changes are coming — and that’s a good thing. Because Linear is step one; after it come Full Leios, UTXO sharding, SNARKs, and data-availability sampling — each step from paper to code may converge again. Hydra went through a similar simplification back then.

But “change” splits two ways. One is subtraction for delivery (Linear here): the functionality stays on its original path, just lands in a smaller first step — that deserves respect. The other is overturning a core commitment: if a simplification ever started relaxing the security model or decentralization, that’s what would genuinely alarm me — and there’s no sign of it.

The last gate is the most telling: every change still has to pass on-chain voting, with the community, SPOs, and exchanges holding veto power. The act of changing itself is reviewed in the open — which reads less like ambition and more like consensus compared to a team deciding behind closed doors.

A guess at the end-state

If this road truly works out, I don’t think the meaning is “the TPS number got prettier” — it’s that the trust assumption of infrastructure gets simpler: an L1 that’s itself fast, steady, and settled clearly means many dApps can run entirely on a single layer without stacking trust; and with on-chain governance, rule changes are public and only altered by consensus. That “single-layer trustworthy + open rules” combo is exactly the shape institutions already know — like a clearing system. Blockchain reaching financial infrastructure needs not just speed but “no need to guess how it changes tomorrow.”

For someone who values stability over noise, every one of these steps has someone accountable behind it. What’s left is to watch whether Dijkstra’s two phases keep arriving on time.

Article Link: