bunch (@octrabunch) header
bunch (@octrabunch) profile picture

head of community @octra

octra
Joined March 2022
Photo by bunch (@octrabunch)Photo by bunch (@octrabunch)Photo by bunch (@octrabunch)Photo by bunch (@octrabunch)Photo by bunch (@octrabunch)Photo by bunch (@octrabunch)Photo by bunch (@octrabunch)Photo by bunch (@octrabunch)Photo by bunch (@octrabunch)Photo by bunch (@octrabunch)
new lambda post 🎉
were quietly finishing up the prototype for octra to work as a priv layer for other networks or apps, test cryptographic proofs and verification took a couple of days (gpu process of building proofs), so we're expecting a couple of updates today. wanted to get definitive confirmation.
1
25
652
earth is ~71% blue. it was always octra coded…
8
2
2
44
1,417
gm octra people
14
46
834
privacy networks are hard as hell. mmost never make it past the whitepaper, or a demo that falls over the second real users show up octra keeps shipping anyway devnet live with encrypted balances, enc/dec and stealth transfers. hfhe proof plumbing landing inside the circles vm. node hardening every week. state sync backed by formal proofs while the rest of the space argues about what "privacy" even means programmable privacy that isn't a mixer and isn't a trusted setup is brutal. most teams quit or pivot these guys just keep pushing commits respect to everyone grinding through the hard parts of fhe instead of tweeting about it
Small but important release: it hasn't been updated since 2024 and we received some criticism around it, as the demo relied on a long deprecated approach that resembled a mock-up. It is now verifiably the current version included in lite_node, and still shows the same performance
4
51
751
HFHE sandbox has been updated to pvac v7 encrypted arithmetic, matrices, vectors, logic gates, fourier transforms and recrypt benchmarks all work locally in your browser with no node, RPC or wallet connection a working demo of real encrypted compute: sandbox.octra.org
16
23
4
86
3,997
octra weekly recap 🔵 → node update: fixed an old bug where one round could get processed twice → OPD: a public log of every protocol change and why. anyone can propose → formal proof that everyone's copy stays in sync after every write → devnet: encrypt, decrypt, stealth transfers and claims all switched on. hfhe proof runs in the vm → alex's loops caught a serious devnet bug on day one, mainnet impact zero → loops won its first bounty from the ethereum foundation: $1,000 hack yourself before others do. links below ↓
octra weekly recap 🔵 → programs now get a compute bill: what ran, what it cost, what got refunded → apps only read circle data that's fully committed → node upgrades set up rust themselves and time out instead of hanging → the appliedml preview now meters heavy math → installer files get checked against a fingerprint list → new octrascan (early alpha): open a program and see what it actually does → alex built a 24/7 ai security loop. agents find the bugs, humans write every fix more next week. same intern, more blue 🔵
6
6
40
1,050
bunch retweeted
the cryptographic proto is fully active on the dev net, the proof form is optimized by 2x (tests can be observed live btw), the entire surface is accessible, enc, dec, stealth txs, claims, etc, and at the VM level, the HFHE proof form (pvac version) is fully operational: devnet.octrascan.io/tx.html?… (you can see an example here and on our dev addr in the history, and the payload is available via rpc and in the raw record) * an important thing before migration, in order to use these features, you will need to upgrade your webcli tomorrow evening – at London, England
4
17
2
73
2,534
bunch retweeted
btw, the day wasn't wasted, we discovered another bottleneck that required special attention, so we decided to deduce it from a proof-of-concept form, rather than simply as an algorithm and separately decided to prove that after any write, the state remains synchronized among all participants, a quorum requires both thresholds (number and weight of participants) to be met the learning context itself, based on moving consensus, doesn't change the vote, lock, or finalization phase. this was a rather specific problem, so we decided to formalize it (sorry for the format - the OPD platform will be available in a couple of days, but for now, we're writing the descriptive part manually), so you can take a look at it here: gitlab.octra.org.uk/octra/li… the new release marker is now available for validators, pls upgrade as usual: sh controls/upgrade.sh sh controls/upgrade.sh --apply sh controls/stat.sh – at London, England
8
11
3
79
3,401
bunch retweeted
Proud to announce that in the early days of external testing, Loops was awarded its first autonomous $1000 bounty by the Ethereum Foundation
Our agentic "Loops" infra is getting better: a serious devnet inflation bug was caught immediately upon activation and with no impact on mainnet. We also ran Loops on friendly privacy/ZK/FHE-adjacent projects, and responsibly disclosed our findings. Write-up coming once safe.
5
7
1
79
1,981
octra protocol document
releasing the latest hand made bits to improve fault tolerance (the upgrade will be required on weekends), as well as a portal with all changes and the proposal log, which stands out for its activation rules, the name has been coined -OPD (octra protocol document), everyone will be able to participate in proposals for changes. OPDs are needed to structure all the changes and separately highlight explanations of why they have been made in the first place, and to propose a discussion.
1
1
1
26
970
bunch retweeted
hack yourself before other do (the policy we ourselves adhere to). fyi: an automated fuzzing search engine based on slop generators (by @octralex) found critical crap in ALL projects we've ever interacted with (since there's already context), some have been fixed, others haven't been revealed yet and may not be, given the mixed reactions. what a time to be alive – at London, England
a new unscheduled release is now available for vals. please upgrade your nodes (a lot has been done on recovery and sync + an old bug from an earlier implementation at the vm level has been fixed) * * the node could process the same epoch twice, and the root at that moment was different, when building and checking proposals at the consensus level, it was specified as 32 raw bytes (the earliest implementation from the end of 2024),and when using the epoch from the head, it was transmitted as a 64-hex string, so the bug was found during pre-release main net testing for mem compatibility of different program envs upgrading is easy - sh controls/upgrade.sh sh controls/upgrade.sh --apply sh controls/stat.sh github.com/octra-labs/lite_n…
5
13
62
1,726
2
6
21
833
bunch retweeted
Our agentic "Loops" infra is getting better: a serious devnet inflation bug was caught immediately upon activation and with no impact on mainnet. We also ran Loops on friendly privacy/ZK/FHE-adjacent projects, and responsibly disclosed our findings. Write-up coming once safe.
a new unscheduled release is now available for vals. please upgrade your nodes (a lot has been done on recovery and sync + an old bug from an earlier implementation at the vm level has been fixed) * * the node could process the same epoch twice, and the root at that moment was different, when building and checking proposals at the consensus level, it was specified as 32 raw bytes (the earliest implementation from the end of 2024),and when using the epoch from the head, it was transmitted as a 64-hex string, so the bug was found during pre-release main net testing for mem compatibility of different program envs upgrading is easy - sh controls/upgrade.sh sh controls/upgrade.sh --apply sh controls/stat.sh github.com/octra-labs/lite_n…
7
7
4
72
4,326
attention, validators upgrade your nodes! thanks for your attention
a new unscheduled release is now available for vals. please upgrade your nodes (a lot has been done on recovery and sync + an old bug from an earlier implementation at the vm level has been fixed) * * the node could process the same epoch twice, and the root at that moment was different, when building and checking proposals at the consensus level, it was specified as 32 raw bytes (the earliest implementation from the end of 2024),and when using the epoch from the head, it was transmitted as a 64-hex string, so the bug was found during pre-release main net testing for mem compatibility of different program envs upgrading is easy - sh controls/upgrade.sh sh controls/upgrade.sh --apply sh controls/stat.sh github.com/octra-labs/lite_n…
1
2
25
896
octra weekly recap 🔵 → programs now get a compute bill: what ran, what it cost, what got refunded → apps only read circle data that's fully committed → node upgrades set up rust themselves and time out instead of hanging → the appliedml preview now meters heavy math → installer files get checked against a fingerprint list → new octrascan (early alpha): open a program and see what it actually does → alex built a 24/7 ai security loop. agents find the bugs, humans write every fix more next week. same intern, more blue 🔵
octra weekly recap 🔵 octrober is here and things are heating up → mainnet is currently being updated → better node syncing + recovery → more work on private program state → better tracking of program compute usage → more AppliedML improvements once mainnet update is complete, all HFHE ops, new aml features, a new program layer + updated circle viewer are coming with it universal encrypted compute is slowly coming together thank u lambda
8
4
1
43
1,949
bunch retweeted
13
23
2
101
4,893
6
5
1
54
2,062
bunch retweeted
Hours after publication, the fleet caught an already patched devnet bug of the same class that we paid a five figure bounty for in June. It was assigned a first ever critical severity and live-tested overnight. 3 months subs paid off tenfold with just one routine finding. Do it!
4
10
80
1,514