<john_r365> sech1: Since the discussions a week or so ago - I had a play turning Opus 4.8 + GPT-5.5 on the Monero codebase to look for potential inflation bugs. After planning out the test areas, I ran it via VPS on a cron to loop through the tests. Same outcome as yourself; nothing found, which is great. And it provides a kind of "framew [... too long, see mrelay.p2pool.observer/e/i8m_7Y4LZGczSU9a ]

<john_r365> I was wondering there's any value running the same process on FCMP++ & Carrot v2.0? I see it's in beta and on stressnet.

<ofrnxmr:xmr.mx> @john_r365: Of course

john_r365 did you make a "handbook" for AI models for the Monero codebase? i.e. ask the AI to map Monero's codebase, describe all its components and their relations, ask questions if anything needs clarification. Further AI reviews benefit greatly when provided this context.

Monero codebase + all the papers available on RingCT/Bulletproofs

<john_r365> The handbook is an interesting idea, I didn't do that, but I'll add it on a future pass now you've mentioned it.

I got carried away by P2Pool vulnerability and subsequent fixes/hardening, but I'll return to auditing Monero codebase next week as well.

<john_r365> Was AI any use in terms of the further hardening aspect?

Yes, I use it to find weak/questionable pieces of code/architecture

<321bob321> Had an idea of using it for spy node finger printing

<321bob321> Feeding it raw logs

<gingeropolous> whats the current number on "spy nodes"? and, do they in effect function as supernodes? (maintain a lot of connections) I guess the question is whether they relay blocks and/or provide bootstrapping service. Well i guess bootstrapping isn't necessary for supernode function.

<gingeropolous> ah bugger. moneronet.info is down

<boog900> @gingeropolous: xmrnetscan.redteam.cash

<gingeropolous> danke

<boog900> > do they in effect function as supernodes

<boog900> They don't, they mostly just listen for inbound connections rather than making a lot of outbound connections

<boog900> they also relay blocks and txs

<boog900> they are just proxying the connections between real nodes

<gingeropolous> but they do make a lot of inbound connections?

<boog900> Well yes but being pedantic they don't make any inbound connections. Other nodes make the connection to them.

<boog900> supernodes are nodes that connect to all nodes themselves

<gingeropolous> context - i'm exploring monerosim's network behavior delta compared to mainnet , and a 15% reachable didn't move the connection-duration closer to mainnets 23 min - it overshot to 150 min. Adding 5 supernodes with max conn 256 didn't budge the overshoot, but it made the sim run faster.

<gingeropolous> "supernodes are nodes that connect to all nodes themselves" > so to mimic a supernode in monerosim i should set the connection limit to 1k+ or unlimited?

<gingeropolous> well wait, a regular node has unlimited incoming...

<gingeropolous> ah right the spynodes main goal is to soak up wallet-rpc conns, or nodes lookin to bootstrap / sync because they presumably need to transact soonish

<boog900> I think the current spy nodes has 2 separate setups, one setup has all IPs pointed at a single node they run. The other has each IP pointed at a real public node.

<boog900> IIRC there are about equal amounts of both types of spy node.

<boog900> so for the first type you would need to make like 25% of the network connect to 1 node

<boog900> or wait 25% of connections 🤔

<thomasbuilds:matrix.org> > <@john_r365> sech1: Since the discussions a week or so ago - I had a play turning Opus 4.8 + GPT-5.5 on the Monero codebase to look for potential inflation bugs. After planning out the test areas, I ran it via VPS on a cron to loop through the tests. Same outcome as yourself; nothing found, which is great. And it provide [... too long, see mrelay.p2pool.observer/e/ldjj9I4LbFlCY1k1 ]

<thomasbuilds:matrix.org> Have to say Monero is well-coded, unlike ZEC which had vulnerabilities surface the moment Opus 4.8 dropped 🤣

<gingeropolous> ah sweet. the 15% did fix the Skellam delta though

<boog900> nice, it matches mainent now?

<gingeropolous> the bots still processing, and im running a reachable % sweep.

<gingeropolous> i love how i now have this incredible tool that would have blown my mind a decade ago, and i find myself impatient waiting for it do things. the coffee and psytrance prolly doesn't help

<gingeropolous> this should be updated soon, i gotta run: github.com/Fountain5405/monerosim/b…/20260620_network_topology_study.md . there's some stuff in there i directed the bot to change and it should push soon