Diehl argues the industry has bifurcated cleanly into stablecoins/custodial exchanges (regulated neobanks with worse UX) and memecoins/prediction markets/onchain casinos (unregulated offshore gambling). The decentralized applications that were supposed to occupy the middle never materialized, and the engineers who would have built them have migrated to AI infrastructure startups.
By submitting Diehl's post and driving it to 312 points, ibobev surfaced the structural-failure thesis to the HN audience. The submission frames Diehl's long-running critique as newly vindicated by the post-ETF, post-stablecoin-regulation landscape.
The editorial reads the HN comment section as the real signal: a Diehl post in 2021 would have drawn furious rebuttals from .eth-handle defenders, but the 2026 thread is dominated by ex-crypto engineers comparing notes on failed L2 teams and asking where the Rust jobs are. The top-voted reply — 'I worked on this for six years and I think he's right' — marks a generational concession from inside the industry.
The synthesis highlights that USDC and USDT now settle roughly $2T/month in 2026, primarily for cross-border B2B payments and remittances — a real product with real demand. But this victory is pyrrhic for the broader crypto thesis: stablecoins are fully custodial and regulated, indistinguishable from a neobank, which means the 'decentralization' premise that justified everything else has been abandoned by the only use case that worked.
Stephen Diehl's latest post, *Crypto in 2026: Oh, This Is the Bad Place*, climbed to 312 on Hacker News with a thesis that lands harder because Diehl has been writing the same obituary, in increasingly precise form, since 2017. The novelty isn't the verdict — Diehl has always thought crypto was a scam wrapped in a distributed systems paper. The novelty is that the comment section finally agrees with him.
Diehl's argument is structural, not moral. He walks through the post-ETF, post-stablecoin-regulation landscape and observes that the industry has split cleanly into two halves. One half is stablecoins and custodial exchanges — fully regulated, fully centralized, indistinguishable from a neobank with worse UX. The other half is memecoins, prediction markets, and onchain casinos — fully unregulated, fully adversarial, indistinguishable from a 2005 offshore poker site. The thing that was supposed to be in the middle — decentralized applications that replaced existing internet infrastructure — never shipped, and the people who were going to ship it have quietly moved to AI infra startups.
The HN thread is the tell. Five years ago a Diehl post would have drawn 800 comments of furious rebuttal from people with .eth handles. The 2026 thread is mostly ex-crypto engineers agreeing, comparing notes on which L2 team folded last, and asking each other where the good Rust jobs are now. The top-voted reply is one sentence: "I worked on this for six years and I think he's right."
The interesting question isn't whether Diehl is correct about crypto. It's why the technical community stopped arguing with him. Three things changed between 2022 and 2026, and none of them were the things crypto advocates predicted would change.
First, stablecoins won, and won so completely that they erased the rest of the use case. USDC and USDT settle roughly $2T/month in 2026, mostly for cross-border B2B payments and remittances. That's a real product. It's also a product that requires exactly zero of the things crypto Twitter argued about for a decade: no consensus mechanism debates, no L2 wars, no MEV, no token. It's a database with a Treasury-bill backing and a regulatory blessing. The killer app for blockchain turned out to be 'a dollar, but programmable,' and once you have that, you don't need the blockchain part to be interesting.
Second, the DeFi composability thesis — that financial primitives would stack like Lego and produce emergent products — quietly died. The 2021 dream was an open, permissionless capital-markets stack. The 2026 reality is that every protocol with real volume is either KYC'd, geofenced, or both, and the composability that remains is mostly used for memecoin launches and sandwich attacks. The actual financial innovation of the cycle — tokenized Treasuries — runs on permissioned chains operated by BlackRock and Franklin Templeton. Composability with what?
Third, and most quietly, the engineering talent migration reversed. From 2017 to 2022, crypto was where the interesting distributed-systems work happened — novel consensus, novel cryptography, novel VM design. Then ZK proving got commoditized, rollups became a solved problem with three winners, and the AI boom started paying $800k for the same skill set with none of the reputational drag. Every former Cosmos, Solana, and Ethereum core engineer I know is now at an inference-infra startup, and they all describe the move the same way: 'I wanted to ship things that people use.'
Diehl's post doesn't say any of this directly — he's too busy enjoying his victory lap — but it's what makes the post resonate. The technical debate is over because the technical people left.
If you're a senior engineer making decisions about what to learn, hire for, or build on in 2026, Diehl's piece is a useful forcing function for a question most people have been avoiding: what's the actual production use case for a blockchain in your system? Be specific. "Programmable dollars for international payouts" is a real answer — Stripe, Visa, and half of fintech now route through stablecoin rails because they're cheaper than SWIFT and faster than ACH. "Provenance for high-value physical goods" is a real, if niche, answer. "Decentralized identity" is not a real answer; it never was.
If the answer is "because the customer asked for it," that's also fine — enterprise sales is enterprise sales. But don't confuse a customer checkbox for a technical bet. The teams that quietly succeeded in this cycle treated blockchain as a payment rail, not a platform: thin integration layer, custodial wallets, off-chain everything that can be off-chain, regulatory posture handled by a licensed partner. The teams that failed treated it as a worldview.
For individual engineers, the resume question is real and worth being honest about. Crypto experience in 2026 reads like Flash experience in 2012: respected as proof you can ship hard distributed systems, but you'd better separate the skills from the ideology on the way out. The signal you want to send is "I built consensus-critical infrastructure under adversarial conditions," not "I believed." Lead with the systems work — the Rust, the cryptography, the high-throughput state machines, the on-call rotations for chains that couldn't go down. Those translate directly to AI inference infra, high-frequency trading, and any backend where correctness under load is non-negotiable. Bury the token launches.
The practical hiring read: a candidate with four years at a serious L1 or L2 team has stronger distributed systems chops than almost any web2 backend hire you'll see, and they're available now at reasonable comp because the market has discounted the whole sector. This is a quiet arbitrage if you can look past the LinkedIn aesthetic.
The Diehl post will get a perfunctory round of "this aged poorly" replies the next time BTC prints a new high, and they'll be wrong, because his thesis was never about price. It was about whether crypto would become a load-bearing piece of internet infrastructure or settle into being a regulated asset class with a casino attached. In 2026 the verdict is in, and it's the second one — which is fine, casinos are a real business, but it's not the revolution anyone signed up for. The interesting work moved. If you're still on the bus, it's worth asking why, and being honest about the answer.
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.