JJJJ

don't get bundled bros stay safe out there

@CloakdDev They are not mutually exclusive, Harmonic requires the bundle control account to be index 0 account of the ComputBudget instruction, Jito has looser restrictions so you can put the dontfront account anywhere, and both will be applied.

@libsvm Would like to use `dontbund1e` but we also send to jito which only support `jitodontfront` afaik Is there a way to have a merged version so we can send to both without having to adjust the account? `jitodontfrontdontbund1e` ?

Introducing Bundle Control Accounts 🛡️ Give your users and applications granular control over how transactions are bundled on Solana. No more frontrunning. No more sandwich attacks. Just simple, prefix-based protection built into Harmonic's block engine.

Simple integration. Powerful protection. Available now. Check the docs to start using Bundle Control Accounts: docs.harmonic.gg/searchers/bund… Building a fairer Solana, one bundle at a time. 🌊

This puts power back in the hands of users and dapps to protect their transactions from MEV exploitation.

new badge who this

Solana Jobs in 2026: - Scheduler Scholar - BisonFi Toxic Taker - Grant Farmer - FORD Adviser - Router Whisperer

when i first started working on solana, anchor programs + solana_program were the bottleneck for execution in standalone benchmarks. extreme bloat. 200k CUs to do some basic operations with anchor, and several thousand for a noop program with solana-program. i pointed this out with solana-nostd-entrypoint and febo fixed this via pinocchio + ptoken. folks have yet to apply this everywhere, but we did our part in giving people the tools to ibrl. this alone brought the spl transfer capacity of the live solana clusters (e.g. mainnet/testnet) from ≈20k TPS to ≈50k tps. then, with the expectation that everyone would write efficient solana programs, i looked into the transaction dispatch overhead in agave. i realized there was some crazy shit happening there. two IBRL prs brought the time it takes to execute a simple SOL transfer from 2.19 micros to 0.43 micros (0.125 us when batched). github.com/anza-xyz/agave… github.com/anza-xyz/agave… with these two changes, the agave VM can execute simple SOL transfers at 8M TPS on a single thread now when batched – 2.3M TPS when not batched.

i'm very happy to share that i have joined the incredible team at @temporal_xyz there is no better team to work on pushing solana forward with and i’m incredibly excited to be part of it thank you for your attention to this matter!

@alphafox @cavemanloverboy

Chicken wing life hack - does this really work?! 😮

@brian_smith_0 @sui414 @harmonic_gg

@sui414 @harmonic_gg Lagging transactions

The median value (since 2026) per Solana slot is 0.019SOL ($1.63) @harmonic_gg median is 0.024~0.026 What is making their slot carrying higher value/fees? 🤔👁️

we do a little scheduling -> improve validator economics -> improve decentralization our fd client (samba) has the waviest rewards & performance of all clients on the market. currently in beta. dm to beta test. Fig 1: cavey is running samba (fd), emerald is running salsa (agave). Fig 2: contrary to popular belief, there is no negative impact on market maker performance when running Harmonic. Harmonic actually slightly positively shifts the 1s markout distribution (and 5s/30s, although 1s most relevant for scheduling) according to last 30D SOL/USDC data across jup, okx, dflow, titan

@mert @cavemanloverboy mev person @cavemanloverboy

@cavemanloverboy lol if the product isn't useful for you then why use it 🤷‍♂️ it's not the best for devs or mev ppl

backpack moved to using orb, which means its time for me to find a new wallet (sorry @mert, i still love you my guy)

@cavemanloverboy @toly Run zlib twice for the best compression per core second performance

@toly toly, respectfully, what are you talking about

optimization is still an area where we can and will see a crazy step function improvement imo its all garbage rn

Calling this Blockworks piece “research” is generous Claiming certain clients harm market structure while showing no statistically significant difference in 5s/HF maker markouts is simply bad analysis. Any trader would know that 30s markouts are useless for this conclusion. They are 19x longer than leader windows and the mean is dominated by long tail noise. Any real researcher would’ve checked distributions, kurtosis, tails, and would've immediately questioned if such a weak and illogical result can support such a strong conclusion. Especially when relevant HF horizons like 5s show nothing.

The trenches are changing