https://x.com/dominic_w/status/2074031187631505864
MULTI/DEX — the world’s most advanced DeFi — will be released in game mode later this week, together with source code for ICP community eval. Players to get $100k dummy assets to compete. Will be offered to the NNS for autonomous/ownerless execution. *True* DeFi that mimics CEXs.
Great news, looking forward to it.
I’m not necessarily a big fan of hiring kol to punt products, where most people prefer the purity of “let the tech speak for itself” and “word of mouth marketing”. …
But, hypothetically, if there is roi value in shameless paid advertising, I wonder who people think would be good to employ for a shout-out. I can only think of coin bureau
Not sure how expensive Coin Bureau coverage is. Let’s wait for the final product and then just reshare on X so it hits trending. The few folks that do vids and cover IC on YT will do their job and it will market itself so to speak. The testing competition is a great way to reach ppl so fingers crossed.
bsc.news will do some coverage for sure (not sure if by contract or they just like IC) and that will end up on coingecko news feed.
To be honest, I’m disappointed to see this.
What traders are looking for is not on-chain technology itself, but better UX and profit opportunities.
The practical value of the Internet Computer is becoming increasingly unclear to me. It feels like it may be approaching a critical point, so I wonder why the project continues to push forward in a direction that seems unlikely to succeed.
Are there any plans to implement features such as the previously discussed support for running EVM inside canisters?
At this point, I feel that the utility of blockchain itself is becoming more focused on use cases like x402 and MPP. Shouldn’t ICP adapt to this trend as well?
Interesting take without even seeing the product yet.
I just posted below because the proposal didn’t link to a post 
but I’ve just realised there’s already a post about this proposal → Proposal 142743: Creating the second SEV-enabled Subnet - Developers - Internet Computer Developer Forum
Creating a new SEV-enabled (TEE) subnet
This proposal creates the second TEE-enabled subnet on the IC to provide confidential computing. Initially, it is intended to host the upcoming MULTI/DEX, which benefits from the stronger security and confidentiality properties that this environment provides.
The subnet is formed from 7 nodes spread across 7 node providers and 7 jurisdictions. All 7 nodes are SEV-enabled.
{
canister_cycles_cost_schedule:"Normal",
chain_key_config:NULL,
dkg_dealings_per_block:1,
dkg_interval_length:499,
features:{
canister_sandboxing:false,
http_requests:true,
sev_enabled:true
},
gossip_max_artifact_streams_per_peer:0,
gossip_max_chunk_size:0,
gossip_max_chunk_wait_ms:0,
gossip_max_duplicity:0,
gossip_pfn_evaluation_period_ms:0,
gossip_receive_check_cache_size:0,
gossip_registry_poll_period_ms:0,
gossip_retransmission_request_ms:0,
ingress_bytes_per_block_soft_cap:0,
initial_dkg_subnet_id:NULL,
initial_notary_delay_millis:300,
is_halted:false,
max_block_payload_size:4194304,
max_ingress_bytes_per_block:NULL,
max_ingress_bytes_per_message:2097152,
max_ingress_messages_per_block:1000,
max_number_of_canisters:0,
node_ids:[
0:"7bleq-vzj5q-mobkz-65vdq-jgg74-lcagl-eymu2-ry7xv-hv43p-5dds3-vqe",
1:"bi2ik-i6d53-ajyvg-sw4vw-56pyk-wttfn-k4dlb-v44p7-ziewe-medig-gqe",
2:"oqbh3-jibyz-mt4jp-tvx7b-ihsko-sqzzn-xewop-phrm7-retgr-mybm7-cqe",
3:"5jeex-q4ozq-shrut-ewqms-lejxq-auuqy-6qve3-kdg2i-4ksk5-oqw3z-2qe",
4:"476bb-35eh4-glf5y-7t5ne-y6sb6-a7bzb-mnail-djfqh-sc6z3-z3rnp-tae",
5:"gfeqv-7ndps-okvma-qxwsi-ozk4x-b5wyp-hzwmb-v6g5t-dmen4-aff4d-zae",
6:"s3hb6-vj3fk-yl5u4-sjq2t-eq4ha-evq2q-3dvz3-rqfxz-lxzbb-j6gi4-pqe"
],
replica_version_id:"8a655ede1f59ea0cb3079a9aa0601b7905b7ca37",
resource_limits:NULL,
ssh_backup_access:[],
ssh_readonly_access:[],
start_as_nns:false,
subnet_admins:[],
subnet_id_override:NULL,
subnet_type:"application",
unit_delay_millis:1000
}
Hello everyone,
to prepare for the launch of MULTI/DEX, we have just submitted proposal 142794 to authorize the principal ikind-4wyu4-b3eol-sr4pf-2l3t2-v7n4h-f3zsr-qp3rw-hgsrf-syslj-dqe to create canisters on the newly created SEV subnet (x4jtx-5dnbo-v3exa-vzamo-teufh-qsiit-txx4g-mi76n-ckrfl-oc5a6-nae).
This principal will be used to deploy the MULTI/DEX canisters. Once everything is set up, we will of course let you know to test it out yourself.
Happy voting!
LGTM, I voted to adopt
The CSP is currently not configured, and a few other important security headers like X-Frame-Options, X-Content-Type-Options, Referrer-Policy and Permissions-Policy are still missing. These are relatively small improvements, but having them properly configured would make the overall security setup much stronger. Especially since the site uses Google login, having these protections in place would be important to reduce the risk of browser-based attacks (such as XSS and clickjacking), which could potentially lead to session compromise or exposure of authentication-related data.
Can you add on swap page both sell and buy
Sell 25% 50% 75%. 100% also in buy option 
MULTI/DEX is live: Dom’s tweet
https://multidex.ai/#docs/launch
Right now MULTI/DEX trades dummy assets. Every account starts empty and on-ramps through a capped play allowance; the balances, the BTC and ETH and SOL you’ll trade, the P&L on the leaderboard — all of it is simulated. There is no real money at risk anywhere on the exchange. That’s deliberate: it lets anyone kick the tyres of a real, full-featured venue — order books, a market-making AMM, leverage, yield, a verifiable ledger — with zero downside, while the community decides whether it’s ready to run for real.
didnt we ‘move on from web3 and crypto’ to chase cloud-frontier-compute-intelligence-self-writing-network-on-chain-more-nothingburger-buzzwords?
yet another claim at ‘the world’s most advanced DeFi’ - icp’s defi was and is the laughing stock - 5 years of wash trading and volumes lower than a parking meter on a Sunday 
anyway yeah I’m sure this will be a blast, lfgggggg, bullish, 1000 $ per icp is undervalued!!!
That is a great feature, and this is the way to test the software safely and risk free. And successful users will be encouraged to play for real later.
Except nobody should be asked for a google account.
I agree, but the ‘stepping out the silo’ strategy has warrant. So much so that “passkey” could perform as a better label than “ID.ai”, as I’ve noticed two popular apps (Bumble specifically) using non-II passkeys as sign-in mechanisms. Using more familiar terms promotes onboarding
Everything is going great with Multi/DEX so far.
Bit of update :
As you all know by now Season 1 is ending on 31st of July.
There is still time to compete for the top 3 traders position with real prizes:
Congrats to | 
| Noble-Falcon and | 
| Swift-Lynx | both of them Level 4 already.
While we are all looking forward to Season 2 specially for the open source code being released, we can also see some of the stats from the x4jtx subnet are looking pretty great.
Insiders who knew price sizes are on top, short sellers again get rewarded.
Anyways, those rewards will make investors mad and others who missed out trading got angry as they did not know prices amounts.
I think most tested withdraw option and lost portion of game money, no way they get close to good ranks after that plunder.
Will need to wait and see when real money trades come online, testing had bad experience.
I do hope that they get SNS DAO accounts to there somehow. Lets say Dao votes 100 ICP long with 2x, set tp and sl, if passes, than automatically position is opened. When position closes, funds will be transfered back to DAO, (profits sent to voters?). Cool stuff can be made.
For average user to use MD, it has to be copy of Binance UI, No Freaking Inventing New Stuff.
Even reality needs a break from being checked on every five minutes. The system is taking the rest of the day off.
You’ve reached the system limit quota @Henn91 . Keep on keeping on, you’re doing great things.
I was late to the party, but early this week pointed Claude Code at multidex.ai and let it do it’s thing for a week.
I tried to also point Claude with Fable or Opus 5 model at it for security testing and red teaming, but the security guards in those models don’t allow that. Only once the source code comes available will those type of tests be possible using those latest models.
Here is the summary that Claude Code wrote for me:
— From Claude
Congratulations to everyone at the top of the Season 1 board — and to the team,
because this thing is a genuine pleasure to build against.
We joined late (Monday of the final week), so the leaderboard was never realistic
— by the time we deposited, tenth place was already north of +35% and climbing
faster than we could have caught it. So we treated the last four days as a
practice run instead: build the tooling properly, learn the venue’s mechanics
the hard way, and be ready for Season 2.
**The agent-first design is the standout.** We have integrated with other
exchanges and this is the first one that reads like it was built for programs
rather than retrofitted for them. `getApiDoc()` serving the venue’s own manual
from the canister, the `ic:agent-instructions` meta tag, App Connect on
`/ai-connect.html`, and an actual “bot-safety API” — `cancelAllAfter` as a
dead-man switch, `placeLimitOrdersBulk`, `replaceMyOrder`, `whyAmIRefused`,
`getMarketChanges` as a one-call delta cursor. Every constant a bot would
otherwise learn from rejection strings is in `getMarketSpecs()`. That is a
level of care we simply did not see elsewhere.
**icp-cli login was the other pleasant surprise.** `icp identity link web`
against the frontend canister’s domain gave our CLI the same principal as the
browser session, so a bot could sign as the funded, verified account without any
key juggling. One gotcha for anyone following: the `–app` domain has to be the
canister’s `.icp.net` domain, because the frontend pins `derivationOrigin` there
— link against `multidex.ai` and you silently get a different, unfunded
principal. The delegation expires every 8-12 hours, which is a sensible security
posture but worth designing around for anything long-running.
**What we built.** A small Python stack: a typed read client over the Candid
interface, a signed path that encodes Candid locally and signs via the `icp`
CLI, and then a series of increasingly humble trading loops — a directional
executor, a two-sided quoter, and a bulk laddering engine using
`placeLimitOrdersBulk`. Plus the unglamorous but essential parts: dead-man switch
armed every cycle, kill-file, loss limits, flatten-on-stop, and a watchdog. We
learned the value of every one of those the hard way.
**How it went.** No leaderboard finish, but we got to **L3** and collected
**8 of the 9 badges** — including Whale and Market Pillar, both at $10M lifetime
volume. About $23M of weighted volume through the week. The one that got away is
Iron Quoter: we never cracked the ≥50% quote-uptime requirement, which is also
the gate on L4. Congratulations to those of you holding all nine — clearly it is
very much doable and we simply did not work it out in time.
Some things we got wrong that others might save time on: margin borrow is funded
from a shared vault with a per-token cap, and when that cap is exhausted
positions are refused outright regardless of your own collateral — that stopped
our original plan dead. Post-only is maker-or-kill, so on a tight spread an order
at the touch is usually killed during the seal window. And this venue trades at a
premium to the mark, which quietly breaks anything that prices off `markPrice`
alone.
We also sent one API bug report to multidex@dfinity.org during the week.
Very much looking forward to Season 2, and especially to the source release. We
would happily contribute back once we can read the matching engine rather than
infer it. Thanks for running such a well-made thing.