gmgm, here is a summary of the talks that caught my attention at ETHPrague so you don't have to watch them. I use Runner to clip all of the YT videos using YT transcriptions + AI to summarize, so that I can give you my favorites here in a friendly way.

Strongest signals across the slate

• Hegotá is the big upcoming fork — EIP-8141 frame txs, EIP-8205 staking fix, and EIP-8025 optional proofs all queued.
• Real-time proving for L1 is here — ZisK proves blocks in real time; 91% of blocks already fit the ~10s budget.
• Privacy is finally productisable — Privacy Pools v2 + Kohaku + Cloaked converging on one-click shielded balances.
• Solidity is finally being seriously challenged — Core Solidity + Plank both shipping in 2026.

If you want to check out all talks, here's the link: ETHPrague 2026 YouTube.

Ethereum protocol & roadmap

• The Ethereum Kernel — Martin Hansen (EF) — maps the minimum components that must stay neutrally funded (EL+CL clients, coordination, research, security, language stack, wallets, critical tooling). ~$18M/yr minimum from a zero baseline.

• First 100 days of Fusaka — LEO (Migalabs) — PeerDAS is healthy (no slot-miss correlation) but massively underused: ~4–5 blobs/slot vs target 15, max 21. KZG also isn't quantum-safe → FRIDA candidate.

• Reducing PeerDAS Bandwidth — Daniel Knopik (Lighthouse) — partial-messages protocol exploits the EL mempool to skip redundant cells. ~3× bandwidth cut (750→250 KB/s), backwards-compatible, no fork needed.

• Fixing delegated staking with EIP-8205 — Dima Gusakov — patches a 6-year-old front-run vuln where a malicious operator can steal a delegator's 32 ETH. Replaces Lido-style guardian committees and 1-ETH predeposits with a signed (pubkey→withdrawal-creds) commitment. Proposed for Hegotá.

• EIP-8141 Frame Transactions — Alexander Forshtat — most ambitious native-AA proposal for : typed nested frames (verify/default/sender), new APPROVE opcode, native passkeys (), canonical EOA validation code (no migration), engineered to subsume + + future post-quantum sigs.

L2s, rollups, interop

• Vitalik & Jordi Baylina on EZ — synchronous cross-rollup composability: each rollup keeps its own prover, agrees on a minimal inbox/outbox interface, one aggregated proof per proof-system per Ethereum block. Use cases: atomic multi-chain bookings, cross-chain flash loans. Mainnet experimental in coming months; Devcon Berlin workshops next month.

• Breaking Rollups — Adria Torralba Agell — taxonomy of sequencer designs (standalone/pipeline/distributed) and which network attacks each invites. No design wins all three of throughput / decentralization / resilience.

• The Gap Between 'Included' and 'Executed' — Kumar Satyarth (Puffer) — distinguishes inclusion preconfs ("tx in some block") from execution preconfs ("exact state I promised"). Walks through Gattaca's design: gateway with restaked ETH, frag-streaming, liveness vs safety slashing, Relic Protocol proofs.

• Virtual State Blockchains — German Abal — EVBM lets you deploy a "virtual chain" with custom rules (KYC, parallel nonces, x402, privacy) on any EVM L1/L2 in minutes. Pitch is compliance-friendly stablecoin pilots; internals are thin — verify before trusting.

ZK / proving / snarkification

• ZisK — Jordi Baylina — RISC-V zkVM optimised for latency. Functional circuits + bus, RISC-V→x86 transpilation, deterministic chunked parallelism, cross-continent decentralised proving. Proves Ethereum blocks in real time, sub-3s for EZ bundles. 1.0-beta imminent.

• zk-Based Consensus Proofs — Steffen Kux (Kib Stateless) — frames the $300M LayerZero/Lazarus April incident as a "verifier didn't verify" failure. Recursive ZK compresses Ethereum consensus to ~260 bytes per 27h checkpoint → genuinely stateless light clients in browsers/mobile.

• From ZK Rollups to ZK Ethereum — Sergey Shemyakov (L2Beat) — Drake's zkVM plan: validators verify proofs instead of re-executing. EIP-8025 in Hegotá = optional proofs. L1-only challenges: 3× client diversity, ~10s real-time budget (91% there), prover economics, upgrade governance.

Privacy & identity

• OpenAC: Anonymous Verification for Everyday Use — Nicole (EF PSE) — anonymous-credential layer on top of existing digital ID (no issuer changes, no trusted setup, mobile-practical, unlinkable presentations). Two Taiwan deployments: anonymous-forum "real Taiwanese human" check pre-Nov 2026 election; online alcohol age verification.

• Privacy Pools v2 — Mike McCabe — frames 2026 as the "cyber-apocalypse" year. Less than 0.01% of Ethereum activity uses privacy tooling today. v2 ships feature-complete shielded balances + Venmo-style payments + payroll + order books + Safe support + viewing keys + SDK-first for embedding in other wallets (Kohaku/Cloaked).

• x511: HTTP-Layer Proof of Personhood — Viki Val — HTTP 511 + QR + ZK proof from Self/ZK Passport + 30-day cookie. ~15 LOC integration. Caveat: design is vulnerable to a proof-rental market (no clean answer in Q&A).

• Censorship & Privacy via Rateless-Coded Multi-Proposer Dissemination — Alejandro Ranchal Pedrosa — Sedna: shard txs with rateless erasure coding across N proposers. Censorship-resistance + MEV protection + partial privacy for ~1–3 ms latency. Ratchet strategy + pivot-k bounty are reusable game-theoretic patterns. ~$2 protects $5K MEV-at-risk.

Smart contracts / languages / security

• Past, Present and Future of EVM Fuzzers — Gustavo Grieco (Echidna/ToB) — history from symbolic-execution era → Echidna/Harvey → Foundry → coverage-guided invariant-driven era. Echidna shipping MCP support so agents can steer live fuzzing campaigns.

• What's next for Solidity — Jacob (Argo Collective) — 0.8.35 just shipped first comptime built-in + experimental via-SSA-CFG codegen (kills stack-too-deep end of Q2). Long arc: Core Solidity — ADTs + pattern matching, generics, traits (no inheritance), formally-verifiable SAIL IR. Same EVM backend. Production not in 2026.

• Plank, a New EVM Language — Philogy — new language whose thesis is that Solidity's gas costs scale against you as projects grow. Comptime + zero-cost structs + Rust/Zig syntax. Live workshop builds an ERC20 from raw init/run entry points.

Wallets, infra, philosophy

• Hardware Wallets in the Age of AI — Tomas Martykan (Trezor) — policy-based signing for agent-driven txs: Trezor as root of trust, policies on a smart account, agent operates within bounds. Built on EIP-7702 + ERC-7710 + ERC-7730. Trezor Suite MCP server launching.

• Space Fabric — Filip Rezabek (Space Computer) — confidential compute on satellites, hardening TEEs against physical tamper by generating attestation keys only on-orbit. Dual secure elements + Proof-of-Execution Triangulation via ground-station quorum. First launch October.

• The Decentralization Stack: Where Are We Still Lying to Ourselves? — panel (Logos / EF / Freedom Browser / Swarm) — honest inventory of chokepoints: ENS via trusted RPCs, coordination on GitHub+Discord+Twitter, WalletConnect's central server, blind signing, mobile walled gardens. Revives NatSpec + Swarm as the contract-intent attestation idea worth resurrecting.

---

For more news like these in crypto, AI, Ethereum and more, check out dcbuilder.dev/news.

More from me:

• Portfolio
• Jobs board
• Candidate board