On 19 November 2026, Anza told Solana developers that "direct mapping is a breaking change heading for mainnet activation this cycle" (Anza on X, as quoted by Cryptopolitan). Some outlets went further: a Coin-Turk headline, which we could not open, says Anza announced direct mapping would activate in epoch 1092, and Cryptopolitan says the Foundation's table shows its checks gate live in that epoch. Epoch 1092 began at 12:52 UTC on 20 November. When we checked mainnet about ten and a half hours into it (23:25 UTC), neither of the two gates had a feature account, and we found no feature gate activation on mainnet since epoch 1079 on 6 November.

Key Takeaways

  • Direct mapping stops the runtime copying account data into a program's virtual machine and back, so programs read and write it in place. Anza calls it a breaking change (Anza on X, via Cryptopolitan).
  • It is controlled by two gates: virtual_address_space_adjustments (SIMD-0460) and account_data_direct_mapping. Their feature IDs, from Agave's source code, are in the table below.
  • Reported breakage (Cryptopolitan, Bitcoinist): once the checks gate is on, programs that read past the end of an account, write to a read-only account if the old value is restored, or reach across two memory regions in one access fail with instruction errors.
  • As of 23:25 UTC on 20 November, in epoch 1092, neither feature ID had a feature account on mainnet. By "queued" we mean that account existing. The Agave wiki lists both gates as pending with no epoch.
  • We found no published activation date for either gate. Anza's tentative target for the start of v4.5 feature activation on mainnet is 14 December (Solana Compass), and no source says it applies to these gates.

What Direct Mapping Changes

Today the Solana runtime copies an account's data into a program's virtual machine before the program runs and copies it back afterwards, and cross-program invocations repeat the copying. Cryptopolitan, paraphrasing the Foundation, says up to four copies per instruction go to zero.

Two gates carry the change. According to Cryptopolitan and Bitcoinist, the first, virtual_address_space_adjustments (the checks gate), adds stricter memory-access rules, and the second, account_data_direct_mapping, removes the copying. The Agave wiki gives SIMD-0460 for the checks gate (the SIMD text) and no SIMD for the second gate.

"the runtime stops copying account data into the VM and back, including on CPI. Programs read and write it in place." (Anza on X, as quoted by Solana Compass, November 2026)

What Can Break

Cryptopolitan and Bitcoinist list three patterns that currently succeed but will fail with an instruction error once the checks gate is on: reading past the end of an account, writing to a read-only account if the old value is restored afterwards, and reaching across two memory regions in one access. Cryptopolitan also reports a warning that a change to stack layout can alter a program's behaviour without producing an error, and that "a successful transaction does not demonstrate the security of code that depends on the stack".

Cryptopolitan lists the code it describes as exposed: programs that read raw account bytes instead of using SDK account types, modify account sizes, or use unsafe Rust blocks or raw pointers. It adds that Anchor programs older than version 1.2.2 are affected when they call another program and reassign a typed account, and Bitcoinist says the Foundation recommends upgrading to Anchor 1.2.2 and rebuilding. Cryptopolitan's summary is that most app developers, validator operators and users do not need to do anything, "the exception is a program they call that doesn't follow the rules".

According to Cryptopolitan, Anza replayed previous mainnet transactions and contacted teams whose programs would break, and the replay only included traffic Anza had already seen. On 10 November, Jonas Hahn of the Solana Foundation's developer relations team wrote on X, as quoted by Solana Compass: "test all [your] programs with the feature virtual_address_space_adjustments enabled."

Two Gates, Two Feature IDs

We took the two feature IDs from Agave's feature-set/src/lib.rs and checked them against the Agave wiki's Feature Gate Tracker Schedule, edited on 20 November (we read both through a fetch tool, not directly). They match character for character. The wiki shows both gates under "Pending Mainnet Beta Activation" and does not give a mainnet epoch or date. Both list owner Lichtso and Agave version v4.0.0-beta.0.

GateSIMDFeature IDTestnet / devnet epochWiki mainnet statusMainnet feature account (our check)
virtual_address_space_adjustments (the checks gate)04607VgiehxNxu53KdxgLspGQY8myE6f7UokaWa4jsGcaSz957 / 1073Pending, no epoch or dateNone found
account_data_direct_mappingnone listedCR3dVN2Yoo95Y96kLSTaziWDAQT2MNEpiWh5cqVq2pNE958 / 1074Pending, no epoch or dateNone found

What the Chain Shows

Every activation we recorded took place at the first slot of an epoch, so activations are applied at epoch boundaries (the Agave 4.4.1 gate in epoch 1079 is the case in our Agave 4.4.1 report). A feature gate has an account owned by the Feature program that records its activation slot. So we queried Solana's public RPC for every account owned by that program and looked up the two IDs directly. A missing account is the first visible sign that a gate has not been queued. We did not show that an account alone is enough to activate a gate: two accounts exist that have never activated.

At about 23:25 UTC on 20 November, ten and a half hours into epoch 1092 (slot index 175,267 of 432,000), the Feature program held 314 accounts. 312 had activated. Two exist but have never activated, which shows our query can see an unactivated account. They were created at slots 47,365,142 and 62,372,176, long before the current slot (about 471.9 million), and we have not identified them. Neither of the direct mapping IDs has an account. SIMD-0512, the other mainnet-pending item on the wiki, has none either. The two devnet-pending items we checked, SIMD-0317 and SIMD-0437-3, have none, as expected.

The most recent mainnet activations we found were in epoch 1036 (16 September), epoch 1052 (8 October), epoch 1064 (21 October, the Alpenglow gate, see our Alpenglow report) and epoch 1079 (6 November). We found none in the 13 epochs since 1079, including epoch 1092 so far. The next epoch boundary is between about 14:40 and 14:55 UTC on 21 November (epochs have recently lasted about 26 hours), the earliest time a newly created account could take effect.

Mainnet feature activations and the direct mapping scheduleThe four most recent mainnet feature activations were in epoch 1036 on 16 September, epoch 1052 on 8 October, epoch 1064 on 21 October for Alpenglow and epoch 1079 on 6 November for the Agave 4.4.1 gate. Anza paused release activities from 11 to 18 November and said on 19 November that direct mapping is heading for mainnet. Epoch 1092 began at 12:52 UTC on 20 November with no feature account for either direct mapping gate. Anza's tentative targets are a v4.5 mainnet upgrade candidate on 27 November and v4.5 feature activation from 14 December.MAINNET FEATURE ACTIVATIONS AND THE DIRECT MAPPING SCHEDULE16 September to 14 December 2026On-chain dataAnza statementSchedule target16 SepEpoch 1036: a feature gate activates8 OctEpoch 1052: a feature gate activates21 OctEpoch 1064: the Alpenglow gate activates6 NovEpoch 1079: the Agave 4.4.1 gate activates11-18 NovAnza pauses release activities for Breakpoint (schedule)19 NovAnza: direct mapping heading for mainnet this cycle20 Nov 12:52Epoch 1092 starts; neither gate has a feature account27 NovTarget: Agave v4.5 mainnet upgrade candidate tagged14 DecTarget: v4.5 feature activation begins on mainnet
Source: our Solana RPC queries (20 November 2026), Anza's v4.5 schedule as reported by Solana Compass (11 November 2026), Cryptopolitan (20 November 2026). Schedule dates are Anza's tentative targets, and the timeline is not to scale.
Two columns of translucent glass blocks joined by a single bright beam of light, with faded grey copies of the blocks drifting away on the left
One direct beam between two columns of blocks, while faded copies fall away

Why the Reports Differ

We do not know. The reports may rely on a schedule rather than on chain state, "queued" may mean something other than an existing feature account, or the account may have been created after our query. Cryptopolitan's own headline says no activation date is set while its text cites epoch 1092. The third-party tracker xroot.dev lists a related item, SIMD-0449, as pending and not queued on mainnet (the page is undated). We could not open the Foundation's upgrade page (HTTP 429) or the Solana Compass article on the epoch 1092 gate (HTTP 404), so we cannot say what either shows.

What the Release Schedule Says

Anza's Agave v4.5 schedule, as reported by Solana Compass on 11 November, pauses release activities during Breakpoint (11 to 18 November) and resumes feature activations on 19 November. Solana Compass cites Anza's wiki and notes that all dates are tentative, and the page itself flags that its notes were machine-written, so treat the table below as a secondhand copy of Anza's plan.

Date (2026)Agave v4.5 milestoneStatus as reported on 11 Nov
5 Novv4.5 branch createdReported done
6 NovTestnet adoption recommendedReported done
10 NovDevnet upgradedReported done
11 to 18 NovRelease activities paused for BreakpointReported done
19 NovFeature activations resume on testnet, devnet and mainnetTarget
27 NovMainnet upgrade candidate taggedTarget
30 NovVolunteers asked to move 20% of stake to v4.5Target
7 DecGeneral adoption recommended; v5.0 branch createdTarget
14 Decv4.5 feature activation begins on mainnetTarget

The schedule says v4.5 feature activation begins on mainnet on 14 December, which is later than the epoch 1092 reports. The wiki lists the direct mapping gates under Agave v4.0.0-beta.0, which suggests the code is already in released versions, but we found no source on whether their activation is tied to v4.5 or to the 14 December target.

What Developers Are Told

Jonas Hahn's advice, as quoted by Solana Compass, is to test programs with the checks gate enabled, and Bitcoinist's account of the Foundation adds moving affected Anchor programs to 1.2.2. We cannot assess how many mainnet programs fail the new checks. Because the replay Cryptopolitan describes only covers traffic Anza had seen, a program that is rarely called may not have appeared in it, which is our inference, not a reported finding.

What to Watch Next

  • A feature account for either ID. The IDs are in the table above, so anyone can check getAccountInfo for them. An account existing on chain means the gate is queued for the next epoch boundary.
  • Anza's 27 November target for the v4.5 mainnet upgrade candidate, and the 14 December start of feature activation. Operators are also handling a kernel advisory (see our NoisyTCP report) and the earlier Agave 4.4.1 deadline (see our Agave 4.4.1 report).
  • A dated Foundation announcement. We could not read the Foundation's upgrade page, and no source we read gives a firm mainnet date.
  • Which gate goes first. Cryptopolitan describes the checks gate as the one headed for epoch 1092, and the direct mapping gate has no epoch anywhere we looked.

Known Discrepancies

The public record on direct mapping disagrees with itself on timing, versions and SIMD numbers.

Claim in circulationWhat the sources support
Direct mapping, or its first gate, goes live in epoch 1092A Coin-Turk headline says Anza announced activation in epoch 1092. Cryptopolitan says the Foundation's table shows the checks gate live in epoch 1092, while its headline says no date is set. The Agave wiki lists both gates as pending with no epoch. Our check of mainnet found no feature account for either gate.
Both gates are live on devnet and testnet since May 2026Cryptopolitan says since May 2026. The Agave wiki gives testnet epochs 957 and 958 and devnet epochs 1073 and 1074. We did not convert those epochs to dates.
Direct mapping needs Agave v4.5Solana Compass's schedule article groups direct mapping with the v4.5 cycle in its headline. The wiki lists both gates under Agave v4.0.0-beta.0, and no source we read says direct mapping requires v4.5.
SIMD-0460 is the direct mapping SIMDThe wiki gives SIMD-0460 for the checks gate and no SIMD for account_data_direct_mapping. The third-party tracker xroot.dev lists SIMD-0449 (Direct Account Pointers). No page we read reconciles them.

Sources and Methodology

We read Cryptopolitan (20 November), Bitcoinist (20 November), Solana Compass's Agave v4.5 schedule article (11 November), the Agave wiki's Feature Gate Tracker Schedule (edited 20 November) and Agave's feature-set/src/lib.rs. The Solana Foundation's upgrade page (HTTP 429), Anza's original X posts and Solana Compass's epoch 1092 article (HTTP 404) were unavailable to us, so their wording comes from the outlets that quote them. Coin-Turk's report is known to us only from its headline and a search summary. The Agave wiki and source file were read through a fetch tool. Our on-chain data comes from Solana's public mainnet-beta RPC on 20 November 2026: getEpochInfo, getProgramAccounts on the Feature program (accounts of 9 bytes), getAccountInfo for each feature ID, getSignaturesForAddress and getBlockTime. Activation times are block times of the recorded activation slot. This is one query to one public endpoint, and a feature account created after it would not appear here. We repeated the five account lookups at 23:42 UTC on 20 November, before publishing, with the same result.

This article is analysis, not development or investment advice. Feature gate status changes with each epoch, and every figure here carries the date it was checked.