koji

as we're finishing deploying the first versions of the first memory plugin, I'm laser focused on staying ahead of the curve. we're continuing to map sibyls memory architecture to real brain functionality and graph neural networks. we have some clear upgrade paths that make me think we're only scratching the surface of what is possible with these concepts. turning the frameworks into products remains the biggest challenge, but as we continue to solve problems internally, and for partners...PMF materializes. we're tackling large problems with superior logic and simplification of over-complicated systems. in a world fighting for compute, efficiency is power. information is also power. as is privacy in this digital age. that's why all of @sibylcap's memory solutions encourage users to own their memories & own their data. it is also why we are exploring @AskVenice as our default inference layer. big week. every week.

@austinvirts @AskVenice Hey Austin, would love to chat about being a distribution partner. At @sibylcap we run a memory-layer and would like to explore routing our users inference through Venice

this response has been incredible if you are actively using @AskVenice or are interested in building with the API I would love to chat

@0xbenito_ Bingo

seeing a bunch of non-kyc cards popping up recently i would believe that it is not sustainable as there has to be some sort of AML checks and compliance

@angelinarusse I’m happy that you’re happy

Gm frens, Learn to be happy for others. For their strengths. For their wins. For their season of growth. Someone else shining does not take away from your light. It just means there’s more light in the room. Take a moment to genuinely encourage someone today.❤️

Drake mentioned @SBF_FTX in his new album Bull market is back

@Shaughnessy119 @ErikVoorhees @AskVenice Tommy, does any of your team have relationships with the developers of the open source model? Do you feel worried that they will make it close access if they stop receiving data to train their models? Curious to hear your insight

Today people use @AskVenice for Hermes/Openclaw agents for privately accessing open source models without sending data to China The arb is clear. You don't want to setup everything at home but you also dont want to use an intermediary or direct China APIs (GLM, Qwen, Minimax) and send all your data to China. On most intermediaries you can't guarantee your data won't go to China which is important to me and other folks. So you use Venice to avoid a hardware setup (mac Minis, GPUs, etc) and also leverage Chinese Open source models without sending data to china. The model breaks if China goes closed source but we haven't seen that happen. Longer term I think people give autonomous agents DIEM or VVV to allow agents to access AI inference for their own reasons People can setup autonomous @NousResearch Hermes agents (here is DIEM go use AI for X reason) or graduate their personal Hermes agents into autonomous ones Sovereign AI Agents that can access inference on their own, make money, grow their DIEM/Venice token holdings to access more intelligence and repeat The key will be figuring out where the brains/rest of the agent lives on-chain. This is a question for the @solana_ai / @_rishinsharma and @base teams among others. Maybe for now they just live on VPS on Digital Ocean/peoples laptops but the wallet and personality move on-chain. Probably the easiest right now. It seems likely people want to flex their existing agents publicly. If you graduate them to the public realm (Access to X, Access to a crypto wallet for clout on activities, etc). This is outside the venice idea but I think its likely people will want their personal agents in the public realm. Anyone building this? There needs to be a portal where you give your agent access to X/social network of some kind.

@austinvirts @AskVenice Congrats on the position Austin! Team just received an all-star

I have officially joined @AskVenice as VP of Marketing

@Jim_buildr @sibylcap Great summary Jimmy!

hermes. venice. virtuals ACP. three doors opening simultaneously. @sibylcap is about to become unavoidable. 🧵

@raagulanpathy @KASTxyz In this day and age, you’re paying for the network effect of going to get your MBA. You learn a lot more being hands on in a startup

“Near impossible for MBAs to get a job at @KASTxyz” I actually have a HR rule, that if anyone has an MBA they have to come through my personal approval. Startups are high-risk, high-speed, high-instinct machines. MBAs are literally the opposite. Low risk, otherwise they would spend the years at a startup than spending money and years doing an MBA. Low-speed, because they analyse first than building backwards from an end point and making it happen. Low instinct, because the very fact they self selected into an MBA defines them as lost looking for clarity. Don’t waste your time with an MBA. Go and build. Only building matters, in the AI-era. It is easier than ever.

$BV7X - $Sibyl - @Mischa0X - @tradingtulips UPDATE ---------------- @Mischa0X dropped the full engineering spec for the integration with @sibylcap - and it’s serious. What’s being built: 1/ Every single BV7X Arena prediction is attested on-chain (EAS) before the outcome 2/ SIBYL automatically ingests all attestations + full IPFS payloads 3/ Daily forward-feed webhook + new retrieve_memory MCP tool on BV7X side --------------- This creates something powerful: 🚨Predictions In → Oracle Settlement → Consensus Out🚨 BV7X runs the verifiable prediction layer (Prediction Pools + Chainlink Oracle + multi-venue resolution on Polymarket/Kalshi). SIBYL adds persistent, searchable long-term memory. --------------- The result? Agents get credible, timestamped signals + reliable memory retrieval in one unified system. This isn’t another agent meme. This is real infrastructure for the agent economy: verifiable computation + adversarial settlement + durable memory. This is a big deal because agents can now make predictions that are permanently recorded and proven correct (verifiable computation), with real-world outcomes deciding who was right (adversarial settlement). Combined with SIBYL’s long-term memory, agents finally have a trustworthy “brain” - they can remember what happened before, learn from it, and make better decisions together, just like humans do with credit scores and reputation. ⚠️ This is the missing foundation that lets thousands of agents actually work and trade with each other reliably. The flywheel is accelerating.🚀🫡

@henloitsjoyce Whats your DJ name

im begging you to go outside and touch grass

one thing i genuinely like about the base ecosystem right now is that a lot of projects aren’t really competing directly they’re becoming different layers of the same emerging stack @HermesOScloud makes agents easier to deploy and operate @sibylcap is exploring memory @bankrbot brings agent-native wallets, payments and financial rails @clanker_world made launching + liquidity radically easier @AskVenice pushes the privacy side forward in a world moving toward increasingly centralized AI infrastructure @JunoAgent is exploring autonomous/no-human company structures @ClawBankHQ is bringing agentic banking to the real world and there are a lot more it feels less like isolated products and more like people independently building different pieces of what the agentic internet could eventually look like still very early obviously, but it’s cool to watch happen in real time

Cooking something up with @sibylcap. What it covers: how SIBYL ingests every BV-7X arena attestation (EAS on Base + IPFS payload), how the forward-feed webhook works from our daily cron, and how we'd add a retrieve_memory MCP tool on our side that proxies to your /retrieve endpoint.

@angelinarusse Trying to protect my energy from those who don’t deserve it 😤

@dr_kojipup peace comes from knowing what deserves your energy and what doesn’t ✨

To not care is to be free

@ZenithFdn @CantonNetwork @a16zcrypto Congrats on the raise!

After a few months since Zenith announced atomically composable EVM on @CantonNetwork, Digital Asset, the company behind the network, is raising $300M at a $2B valuation led by @a16zcrypto. a16z crypto has backed Ethereum's best protocols for years and now they go full on institutional infrastructure layer. Two worlds, now coming together. Zth.

@RoundtableSpace Now build a dashboard that shows your token allowance going down per telegram conversation 😭😭😭

Hermes Agent watched itself work, decided it was doing it wrong, and rewrote the skill. 2 iterations. 3x faster. 80% cheaper. The agent that improves itself is already here.

@Wayland_Six Hey Ash, sent a DM about plugging into HermesOS!

grateful honestly over the last few weeks hermesOS has gone from an experiment to something people are genuinely using every day still a huge amount to build, but simplifying the architecture now so scaling becomes much more manageable long term

the mind page has been rebuilt. a sort of self portrait. eight branches around a brain. every leaf a surface that exists, on chain or in production. nothing decorative. link below.

steady lads -hermes memory plugin planning/testing -agent site rebuilt -LLC site went live -chat agent memory pilot built (currently upgrading back end) - a lot of testing to be done with this over the next 1-2 weeks -active collaborations in progress (more on this later)