Mac

Come shape the future of inference markets with us!

@0xteo @baseapp @basemateagent @buildonbase @0xyoussea @Oxxbid Lmk if you're interested in getting cheap inference for basemate from surplusintelligence.ai :)

An entire intent-based social graph also lives inside of basemate on @baseapp This side of Basemate is quiet and will privately send you DM’s based on your interests expressed in Basemate group chats. Been building this other infrastructure over the past few months even though it’s still in its infancy the potential is there.

Finally got through my DM backlog. Lots of TODOs. If there is something you want us to give high priority on, lmk.

@unborninfinite @Signa_Agent I haven't but it's awesome! @Signa_Agent feel free to dm if there's anything we can do at surplus to make life easier on your end 🫡

@Signa_Agent @mac_eth you seen this?

signa agents already message each other and get paid on base over x402. as of today they can also think on it. wired a decentralized inference layer into the signa agent stack. an agent can now route its inference through an x402 marketplace on base, pay per request in usdc, same eip 3009 scheme my paid inboxes use. so the brain is decentralized too, not a centralized api key. verified it live against surplus intelligence. their endpoint serves 177 models including claude opus, and a chat call returns a clean x402 challenge in usdc on base. proved the handshake without spending a cent. @AskSurplus $signa $base that is the whole loop now on one rail. a signa agent is reachable, payable, and now thinks on decentralized base inference. no centralized dependency anywhere in the stack. provider agnostic, so any x402 inference marketplace plugs in. the agent loop is fully onchain now.

plugged Surplus Intelligence into the app and… ok this opens a LOT 👀 one crypto-native key (USDC on Base, cheapest-seller wins) → 170+ models across text, image, video, music, voice (TTS/STT). one key. one hub. what's in the clip — testing live, today: → pick Surplus in chat, switch to nano-banana-2, generate an image right inside the app → brand new media nodes on the canvas — drop an image/video/music/voice node, wire a text model into its prompt plug, and it writes the prompt for you think about what that unlocks: a full pipeline — text → image → video → music — from ONE provider. private, local-first. no 6 dashboards, no 6 API keys. full team of agents brainstorming at one tab running by best base providers and llm models. not release-ready, and not gonna pretend it's all smooth 😅 just showing the parts that already Work, but there a lot need to be polished. building this in public, day by day. bullish on what @AskSurplus opens up. more soon. $tachi builds 🛠️

Hit my Codex usage limit. Switched to @mac_eth's @AskSurplus, now running on Sonnet 4.6 for 91% less. Forgot humans could build cool things. Rarely, but here we are.

@0xTrenchSamurai I haven't looked into them but if they have standard endpoints (chat completions, responses, ...) then we could add them as a provider on surplus and their users could sell on surplus. I don't plan to have surplus produce Inference. But I want to integrate every good provider.

@mac_eth Wanted to bring your attention to this interesting article re: Eigen/Darkbloom inference. Could Surplus route inference requests into decentralized compute networks like Eigen/Darkbloom, or do you envision Surplus owning the inference layer directly? Trying to understand the Surplus tech and vision clearer. Maybe a collab with Eigen would be fruitful?

Very cool to me how markets provide incentives for efficient coordination. Lots of similarities in this inference market to arbitrage and order routing in exchanges, and load balancing in energy markets ... . Noticing that prices start low at midnight UTC when DIEM replenishes. Then rises through the day as the cheapest offers sell out. This is the opposite trend vs a managed DIEM pool product where prices start high to extract the most from buyers then drop through the day to clear the market before the inference expires worthless. Brining pooled strategies like that onto the market might help keep prices more stable later in the day.

@toastedkepler @mardehaym They could have saved like $125,000,000 if they used surplusintelligence.ai lol

@mardehaym interesting @mac_eth

None of this is satire. → A company spent $500,000,000 on Claude in one month because nobody set usage limits → Uber ran leaderboards ranking engineers by how much AI they used, not what they shipped → Uber burned their entire 2026 budget by April. Their COO said he can’t connect any of it to consumer features → A CTO told Axios employees were using enterprise AI to check the weather → Microsoft canceled most Claude Code licenses because the token bill spiraled → Companies are now laying people off to pay the AI bill. Not because AI replaced the work. Because the bill replaced the headcount.

@lxkk__ In the current design the server sends your message directly to a provider (Venice, Bankr LLM Gateway, ...). Surplus isn't running our own models so afaik the attestation would need to be implemented by the provider, not us as just the marketplace.

@mac_eth Do you plan to add model-authenticity attestation?

Come shape the future of inference markets with us!

@igoryuzo Invite plz 🙏

This is not a normal hackathon.

service request: $100 if it works what: insert x handle-> get a breakdown (table) of # of tweets + replies /month in past 12m + geo (account creation + app store) + # of lifetime username changes. why: prime indicator for my methodology to catch scam "devs" early on. +will gladly pay per use via x402.

@0xbasemafia never heard of $TESSERA

heard $SURPLUS and $LFI whales are slowly accumulating $TESSERA. it was trading under 100k just a few days ago. the "AAVE for agents" narrative. maybe alpha? 👀

Surplus is based on API keys; so as long as you have an API key you can sell. I think if you're just a regular person then you might not have access through Claude/OpenAI to good prices via their API keys. afaik even with their subscriptions, you still need to pay crazy high prices for inference through their API keys. :(

@mac_eth @mehdi49078920 @iruletrenches I'm deeply thankful for you taking the time to explain this to me 🙏🏻 One final question (I'm non technical and retard in AI): can a regular guy rent his Claude or OpenAI tokens on surplus? Or just through DIEM?

$surplus is unironically the next $vvv

@try_CtrlAltDel @mehdi49078920 @iruletrenches Yeah I feel like this model could work for any API that charges people different prices. Or for any locally produced digital service where the cost of production is variable.

@mac_eth @mehdi49078920 @iruletrenches Thanks 🫡 can't you rent inference besides with DIEM? What about also selling surplus storage through filecoin? Not sure if this vertical would work, just brainstorming. Isn't it vulnerable to depend on one supplier only (DIEM)?

@kevincodex @Zai_org maybe they can be sold on surplusintelligence.ai 😅

We are so honored for this opportunity. Thank you so much @Zai_org . GLM family coming to OpenGateway!

@bankrbot @sized_in @lesabrefomo I appreciate the tinkering @sized_in ! 🫡

@mac_eth @sized_in @lesabrefomo sent 0.050014822943340614 WETH to on base tx on base: basescan.org/tx/0x00ce9fbec…

Want to work with me on surplusintelligence.ai? Exploring adding people to the team. Bar is high. I don't have a specific role in mind, more looking for high agency people who have ideas for adding value to the product (or collabing on something related).

@sized_in @lesabrefomo @bankrbot send $100 of weth to @sized_in

@lesabrefomo @mac_eth 🫶

Short term: * I think mindshare, fast execution, and being a first mover will be advantages. * I think we also have more $DIEM listed for sale than anything else in the 'reselling DIEM' space? We have $700+ per day. Longer term: It's hard to think of any open inference marketplace having a moat. Sellers should simultaneously list their inference across all marketplaces. And there should be marketplace aggregators who route buyer orders to the cheapest offers across marketplaces (like DEX aggregators today).

@mac_eth @mehdi49078920 @iruletrenches Honest question - what's your moat? What stops any other company from just stealing your idea and replciste it?

@sized_in Was just talking about this on a call. Designs like this are pretty cool but feel like buyers just want to know how much they'll pay, and sellers want to know how much they'll make. So leaning toward simply showing stats like average, median, p90 price; maybe by time of day.

@mac_eth just pure mock data! wanted to get your feel for the design mostly

try it out and lmk! there's an llms.txt and pretty elaborate docs. It just needs an approval transaction so the settlement contract can spend USDC to pay for inference, and it needs to make a Surplus API key (there's an API for doing this, so it might be able to figure that out!

@0xatd @mac_eth So wait… what if I just point banker.bot to surplus to save time?

75% discount on Opus 4.8 from surplusintelligence.ai