Valy

Amadeus Mindshare Stream x.com/i/broadcasts/1…

Nova is how we make agents operational, not just conversational on @amadeusprotocol Most “agent stacks” break the moment you move from a demo to production: > workflows become brittle > tool calls drift > behavior changes without a clear reason > nobody can replay what happened Nova turns an intent (“run this strategy”, “rebalance this portfolio”, “route this workflow”) into a structured execution plan: > sequence the steps > call the right tools / contracts > manage state + permissions > settle payments (agent-to-agent / pay-per-call) > record upgrades so evolution is traceable For example, not every task needs frontier models and can be solved cost efficiently. When agents handle real value, the orchestration layer becomes the product: it’s what makes execution predictable, debuggable, and scalable. Nova is our bet that the next generation of crypto apps won’t be “apps” at all. They’ll be long-lived agents that coordinate, transact, and improve… with an orchestration layer designed for that from day one.

looks like anybody now can integrate the agent’s identity through mcp

@paulhsu @NEARProtocol we are building layer 1 for private long-running agents, would be amazing to meet at a #NEARCON

Looking forward to speaking at #NEARCON next week in San Francisco: “How Founders Build and Scale AI-Native Crypto Products” As an early-stage investor focused on Web3 x AI, I spend my time with founders navigating a new design frontier. @NEARProtocol

@ai @openclaw thanks

If you are running into the "Provider [X] is in cooldown (all profiles unavailable) (rate_limit)" issue on @openclaw, run ```openclaw models status``` and you can see how long before you are in timeout for.

It's happening Now that we acquired Bitte, we are upgrading its runtime with an intelligent orchestration architecture. Instead of routing every query through the same expensive frontier model, @amadeusprotocol classifies each task up front, assigns it to the right tier, and puts a small, open, cheap orchestrator (e.g. 8B parameters) in control — with consultant models available for planning and heavy reasoning when needed. The evidence: NVIDIA's recent ToolOrchestra research shows an 8B orchestrator outperforming GPT-5 on Humanity's Last Exam (37.1% vs. 35.1%) while spending a fraction per query. We are applying the same. What we expect in production: 3–5× better cost-effectiveness versus today’s monolithic runtime, depending on task complexity. We keep cooking. Come and build your agents with AMA

@hthieblot l1 for private long running ai agents

I'm investing up to 250k first checks in teams building: - robotics, drones, space - crypto - applied ai/ml - ar/vr - manufacturing, logistics DMs always open. Tell me what you're building!

good or bad 🤔

@CanonicalCrypto gm ✨

gm

@OpenVC_ how true is this, NDA’s also can make an investor liable which is an unnecessary risk

Should founders ask investors for a NDA before sharing their decks? The answer is no. Here are the 6 reasons why, and how you can actually protect your startup without a NDA. openvc.app/blog/nda 🔒

@OpenVC_ nice episode!

On this episode of The OpenVC podcast we talk to Francis Santora. Francis has been a full time angel investor for three years, and in that short time, he has had 2 investments reach $10m in ARR. Francis uses OpenVC.app to source high quality investment opportunities.

@AriannaSimpson congratulations!

Some news! After 6 incredible years, I’m going to be transitioning out of a16z. I’m starting a fund of my own to do what I love most, which is investing in great founders as early as I can find them, with a broader aperture across the many verticals where great companies are being built today. I learned a huge amount during my time at a16z. @cdixon is widely known as a legendary investor and having the opportunity to work closely with him for the past 6 years has been an honor. I am extremely grateful for his mentorship, the opportunities he gave me here, and for the capital and responsibilities he entrusted me with. His frameworks will shape how I think about investments for the rest of my career. When I joined, the crypto vertical was 7 people (it’s now north of 80), and the firm, while already successful, was nowhere near the scale or scope it has today. At the time I thought the firm’s moves to dominate the industry had mostly unfolded, but I underestimated how much the lead could widen in a few short years. @bhorowitz and @pmarca have built an institution, and I am glad to have had the chance to play a small role on the team. I’m sure I’ll be looking back 6 years from now and see the firm in a position  that’s  hard to even imagine from today’s vantage point. Most of the best people I’ve worked with in my career to date have been at a16z – there are too many to name. I’m very grateful to have worked with so many incredible folks here, and I know that I am leaving the crypto team and the investing practice in extremely capable hands with @cdixon, @alive_eth, and @guywuolletjr. I’m also really going to miss working with @jasonrothenal and @eddylazzarin every day in Menlo Park — I’ve learned so much from both of them. I’m very proud of the work I did here, and most importantly, of the founders I had the privilege of working with. They are the reason why I love being an investor. Sometimes it takes dozens of meetings, but every time you find a star, it makes you fall in love with the job all over again. a16z’s passion for and commitment to founders and their companies is what made me love this job in particular. I’m excited to keep doing that in my next chapter — and if you’re building, I’d love to meet!

@topology_vc @openmind_agi congratulations!

Congratulations to the @openmind_agi team on their Series A fundraise. Topology is proud to have led OpenMind's seed fundraising in 2024. The intersection of robotics and AI is just getting started.

ai-native l1, check out our litepaper

@airstreet AI is clearly the future!

Epoch 2

@decasonic 100% agree, trust is key for new emerging AI-focused L1/2s

The rise of AI-native social layers on L2 networks signals a new frontier: agents transacting, coordinating, and creating value onchain without human intermediaries. The opportunity is building verifiable, secure agent economies that can scale. Adoption will favor ecosystems that solve trust at the protocol level.

@realdavidawad yeah 😂

> the tests aren't covering 20% of the code so it fails the threshold > tell the bot to fix the problem > bot lowers the coverage threshold These things really are junior software engineers.

@jayendra_jog the real setups use both confidential CPU and GPU, everybody who deploys confidential CPU with exposed non-secure GPU must understand the implications

huge blow for TEE believers the dominant approach to splitting LLM inference between TEEs and untrusted GPUs is fundamentally broken big breakthrough from a paper that came out this morning

@base @stripe I think this will drive massive adoption

BREAKING: @stripe just launched x402 payments on Base, letting developers charge AI agents directly using USDC.