Platåberget, the public Ethereum testnet built to break Glamsterdam, activated its Gloas fork at epoch 1536 on 2026-08-20 at 07:50:24 UTC. That is the advertised clock, not a rumor. The explorer the testnet site points to already marks Gloas as active. Finality has not crossed that fork yet, and missed blocks jumped in the next epochs. This is a lab result, not a mainnet green light.
Ethereum (ETH) last printed at $2,249.97, up 17.5% over 24 hours, on the site CoinGecko snapshot fetched at 2026-08-20 08:05:30 UTC. That tape is the live chain. Platåberget is the dress rehearsal. Do not mix them up.
I wanted the fork to fire on time. I also wanted the first hours to look ugly if the new rules were real. So far, both happened.
What happened
The Ethereum Foundation announced Platåberget on 2026-08-17 as the public Gloas testnet for Glamsterdam, including enshrined proposer-builder separation (ePBS: the protocol, not a sidecar, separates block proposing from block building) and other upgrade work. The testnet site still says the validator set is open and the network is active. That badge was already there before today’s fork.
What changed this morning is the fork schedule, not the homepage slogan. Platåberget’s published config lists Glamsterdam (Amsterdam / Gloas) at timestamp 1787212224, which is 2026-08-20 07:50:24 UTC, epoch 1536. The Dora explorer API for the same network, read at 2026-08-20 08:15:52 UTC, reports Gloas as an active consensus fork at that epoch, version 0x80733183. Current slot was 49279, current epoch 1539 (the epoch had just filled). Active validators: 84,264.
The same Dora epochs feed is the useful part. Epoch 1534 is finalized. Epochs 1535 through 1539 are not. Justified epoch sat at 1535. In proof-of-stake language, the chain is still producing heads, but it has not locked the Gloas epoch yet.
Block production around the cut is uneven. Epoch 1535 (pre-fork) shows 33 proposed blocks and 0 missed, with sync participation about 33%. Fork epoch 1536 shows 31 proposed, 0 missed, sync participation about 29.5%, and 2 orphaned blocks. Then the next completed windows get rough: epoch 1537 has 12 proposed and 20 missed (sync about 4.8%); epoch 1538 has 9 proposed and 23 missed (sync about 4.2%); epoch 1539, still wrapping when we pulled the feed, shows 13 proposed and 20 missed (sync about 5.6%). Those are explorer counts, timestamped to that pull. They are not a mainnet outage.
A later scheduled event on the same testnet config is a 200 million gas limit (EIP-8261) at timestamp 1787223744, 2026-08-20 11:02:24 UTC, epoch 1566. That print has not happened yet at the time of this reaction.
Context
We already wrote the announcement: Platåberget exists to find Glamsterdam breakage before public testnets and mainnet. Today’s activation is the clock that piece was waiting on. It does not retire that caution. It starts the measurement window.
Gloas is the consensus name that pairs with the execution “Amsterdam” label. Together they are Glamsterdam. ePBS is the headline mechanic: builders and proposers stop sharing one job description. If that wiring is wrong, you see it first as missed slots, weak sync, and finality that refuses to walk forward. That is why a dedicated public testnet exists instead of a blog diagram.
Mainnet gas fees are a different dataset. DefiLlama’s Ethereum fees overview, read at 2026-08-20 08:14:00 UTC, shows about $9.13 million in fees over the last 24 hours, up 27.19% versus the prior day. Traders paying those bills are not on Platåberget. The fee print only tells you the live chain is busy while the lab chain takes the upgrade punch. Show me the fee on mainnet, and show me the missed slots on the testnet. Do not use one as a proxy for the other.
The Ethereum history page is the longer arc: merges, fee-market changes, and the habit of testing in public. Platåberget sits in that habit. It is still a testnet. Deposits, faucets, and throwaway validators are the point. Nobody should treat epoch 1536 as an ETH listing event or a proof-of-stake mainnet milestone.
Client software is part of the same desk. Last week’s Lighthouse v8.2.2 note was about patching beacon nodes on the networks people already run. Today’s story is the Glamsterdam lab fork actually flipping. Operators watching both should keep them in separate tabs.
Our read
My stance: Gloas is live on Platåberget on the advertised epoch, and that is still a break test. The first 25 minutes after 07:50:24 UTC look like a fork that fired, then a chain that struggled to keep a full slot schedule. I would rather see that mess on this network than a quiet chart that hides the new rules.
Do not read “Gloas active: true” as “ePBS is ready for mainnet.” An explorer flag can turn on while justified/finalized checkpoints lag the fork by several epochs. Finality last sat at epoch 1534, two epochs before Gloas, with justification at 1535, one epoch before. Head is already at 1539. That gap is the story for this morning, not the price of ETH.
Falsifiable claim: by 2026-08-21 12:00 UTC, Dora’s Platåberget epochs feed will show finalized_epoch at or above 1536, and at least one fully complete post-Gloas epoch (1537 or later) with missed_blocks at or below 4. If both prints appear, the “fork first, messy hours” read is wrong and I will update it. If ethPandaOps or the Ethereum Foundation instead publish that epoch 1536 was not Gloas, or that the fork was rolled back, the activation read is wrong.
What to watch next
Watch Dora finality. If finalized_epoch stays at 1534 while the head keeps climbing through Gloas epochs, the lab is doing its job the hard way. If finality walks through 1536 this afternoon, the first-hour mess was a blip, not a stall.
Watch epoch 1566 at 2026-08-20 11:02:24 UTC for the scheduled 200 million gas-limit event. A second stressor a few hours after Gloas is a better test than a single fork epoch.
Watch whether missed_blocks fade on later epochs or stay in the 20s. And keep the live ETH tape in its own frame: snapshot prices here, testnet slot counts there. For a refresher on that split, see how to read a crypto market snapshot. More Ethereum desk copy lives under Ethereum news, including our Platåberget announcement reaction.
Primary sources for this note: the Platåberget testnet site (fork schedule and network copy), the Dora network overview API plus its epochs feed, the Ethereum Foundation’s Platåberget announcement, and the DefiLlama Ethereum fees overview.