Blog
Notes on Solana shreds
Field notes on receiving and reading raw shreds.
- Packets are not data: what happens after a shred arrivesA raw shred is a fragment of a block, not a record you can store. Here are the six stages that turn those fragments into rows, and the outside state each stage needs.November 18, 2026Blog
- The lookup table problem: decoding v0 transactions before anyone confirms themA v0 transaction names many of its accounts by table position, not by key. That one design choice is why your decoded account lists come out short, and why the fix is a cache with rules rather than a single RPC call.November 18, 2026Blog
- What your decoder cannot see: the blind spots of shred dataA shred carries the signed transaction and nothing the runtime produced afterwards. Here is the field-by-field list of what is missing from raw data, why, and what you can still work out.November 18, 2026Blog
- One transaction, many instructions: reading a real bot transaction in orderReal traffic rarely holds one instruction per transaction. Here is how to read a composed transaction from top to bottom and store it as one action with its setup steps attached.November 18, 2026Blog
- Alpenglow, seen from an indexer's chairAlpenglow changed how Solana agrees on blocks, and an indexer feels it in three places: its status column, its vote handling and its block parser. Here is what to change, and what you can leave alone.November 18, 2026Blog
- The votes left the block: what that does to your streamUnder Alpenglow, vote transactions stopped appearing in Solana blocks. Code that filtered them is now idle, and any count that included them changed meaning at one slot.November 18, 2026Blog
- Five slots a second: designing writes that keep paceShorter slots doubled the rate at which a per-slot writer has to commit. Here is how to choose between batching by slot and batching by time, and how to size the write path on your own hardware.November 18, 2026Blog
- Write it twice, store it once: idempotent ingestion for chain dataA shred pipeline sees the same transaction more than once: after restarts, replays and late recovery. Choose keys and upserts so that every repeat lands on the row it already wrote.November 18, 2026Blog
- Provisional rows: storing what you saw before the chain agreedData decoded from shreds is early by design, so it has to be stored as a claim, not a fact. Here is a small status model, the events that move a row through it, and the queries that keep early rows away from readers who expect final data.November 18, 2026Blog
- Landed but failed: why a transaction in a block may have done nothingA transaction can be packed into a block, pay its fee and still change nothing else. Raw shreds show it exactly like a successful one. Here is why that happens, how often it happened in one busy program on the day we checked, and how to keep it out of your figures.November 18, 2026Blog
- Filter early, filter cheap: placing the program filter in your pipelineA raw feed hands you every transaction, and most pipelines keep a small share of them. Where you discard the rest decides both your CPU bill and what you silently miss.November 18, 2026Blog
- Your decoder will meet garbage: fuzzing a Solana transaction parserA transaction parser that has only seen well-formed input will meet truncated, oversized and nonsensical bytes sooner or later. Fuzzing finds those inputs on your schedule instead of in production.November 18, 2026Blog
- Golden fixtures: regression tests that catch a broken decoderA decoder change that silently alters one field in one program's output is the hardest bug to notice. Golden fixtures make every such change show up as a diff before it ships.November 18, 2026Blog
- Borsh pitfalls that break Solana decodersBorsh errors in a Solana decoder are rarely bugs in Borsh. They are layout mismatches, and each error message points at a specific one.November 18, 2026Blog
- IDL drift: when a program upgrade breaks your decoderPrograms you decode will change. A decoder that applies yesterday's layout to today's bytes rarely crashes; it writes wrong rows. Here is a policy that turns drift into an alert.November 18, 2026Blog
- Watching for program upgrades as they happenWhen a program you depend on is redeployed, its upgrade is an ordinary instruction to the upgradeable loader. Here is how to recognise it, and its authority changes, in raw transactions.November 18, 2026Blog
- Event schemas that last: modelling decoded Solana activityThe decoded events you publish today will be read by code written next year, after the programs behind them have changed. Separate what you saw from what you concluded, and version the conclusions.November 18, 2026Blog
- Backfill and live: joining history to a stream without gapsHistory comes from RPC, the present comes from raw shreds, and the seam between them is where rows go missing or get counted twice. Here is how to place the seam and prove it holds.November 18, 2026Blog
- A Pump.fun launch, byte by byteOne real token launch, taken apart: five instructions, sixteen accounts in the create call, and 164 bytes of create data decoded against the program's official IDL.November 18, 2026Blog
- Who made this swap? Attributing Solana trades to wallets and routersFrom raw transactions you can tell which program a user called and who signed, but not which pools an aggregator touched underneath. Here are the rules that hold, and where they stop.November 18, 2026Blog
- Counting DEX volume before the chain confirms itA swap instruction states how much a trader is willing to put in or how little they will accept out. That is enough for a live volume figure, as long as you label it and correct it once results arrive.November 18, 2026Blog
- Wallet alerts without false alarmsMost false wallet alerts are not decoder bugs. They are real transactions that failed, never landed, arrived twice, or were never meant to matter. Here is each cause with its fix.November 18, 2026Blog
- What decoding costs you in CPU, and how to measure itThere is no honest single number for how many cores a Solana decoder needs. There is a repeatable way to find yours: time each stage on recorded traffic, then size for the busiest slots.November 18, 2026Blog
- Zero copy parsing for Solana transactionsA transaction arrives as bytes you already hold. A borrowed view reads signatures, keys and instructions in place, and hands out slices instead of fresh vectors. Here is a tested parser and the cases where you still need to copy.November 18, 2026Blog
- Spotting new accounts in the raw transaction streamA handful of instructions create Solana accounts, and each says something different about whether the account is really new. Here is the list, what each proves, and how to confirm it afterwards.November 18, 2026Blog
- Observability for decoders: the metrics that catch silent failureA decoder rarely crashes. It keeps running while it drops a stage's output, misreads a program after an upgrade, or falls a few slots behind. These are the metrics that notice.November 18, 2026Blog
- Order inside a slot: what entry order tells youA Solana block is an ordered list of entries, and each entry is an ordered list of transactions. Flatten the two and you have the block's transaction order, but only once you hold every piece of the slot.November 18, 2026Blog
- A research dataset from raw shreds: what to keep and howA dataset is only reusable if someone else can tell what it contains, how it was captured, and where it has holes. Here is what to record from raw shreds, in which formats, and what to write down next to it.November 18, 2026Blog
- Token 2022 surprises for decodersToken 2022 keeps the familiar instruction shapes, then adds fees that shrink what arrives, hooks that bring their own accounts, and amounts you cannot read at all. Here is what each means for a decoder.November 18, 2026Blog
- Buying shreds as a developer: what you actually needA raw shred feed is a stream of UDP packets and nothing more. Here is the checklist a developer should run through before ordering one: what you receive, what you build, what to send us and how the first day goes.November 18, 2026Blog