Spice up your Telegram feed anon: 👉t.me/telescopetrgc ✨

Seedream 5.0 Lite is on Prodia 🚀 Here's what makes it special! * Face and identity consistency, even when props and backgrounds change * Smart scene adaptation so you can move people and objects to new locations, even groups * Multi-image blending lets you combine elements from different images * Accurate text transfer so you can move text to new backgrounds and items, like merch

Voice AI is projected to surpass $100B by 2030. 🤖 Not because it’s trending but because it’s becoming foundational. Every assistant, every call center agent, every robot, every autonomous system that interacts with humans needs to understand speech. Not just words but tone, context, intention. The demand is obvious. What’s less obvious is the constraint. Most voice models are trained on controlled datasets. Clean recordings. Limited speaker pools. Narrow accent distributions. A handful of dominant languages overrepresented again and again. That works. Until you deploy globally. Because the real world doesn’t speak in one accent. It speaks Spanish in Bogotá and Spanish in Madrid and they don’t sound the same. It speaks English in Lagos, London, and Manila. All different. It blends dialects. It carries cultural rhythm. It shifts tone depending on context. You can’t manufacture that diversity in a lab. You can’t simulate millions of speakers across 180+ countries with authentic linguistic variation and lived context. And that’s where the gap emerges. The next generation of voice AI won’t win because it trained on more of the same. It will win because it trained on broader, richer, more representative speech. High-quality. Clean. Consent-driven. But globally diverse. Multilingual, accent-rich, real-world speech data at scale is still scarce. That’s our opportunity. We’re building the supply for a demand that is exploding 🤫

POV: you're a professional AI character designer and you wake up one morning realizing AI growth is inevitable @xAnnyBuh: 'it's imperative we stay on the good side of AI' ‼️‼️ @IRA_therabbit : 👁️👄👁️ peak Hot Girls Vibe Design content 👇 next ep: Thursday 10 AM CET

Infrastructure isn’t just steel, fiber, and spectrum anymore. With Althea + Liquid Infrastructure, networks can track usage in real time and move payments just as fast as they move data. 1/3

✨ New look: The agentic era is here! As this era unfolds, agents will observe, decide, transact, and execute tasks with real value on the line. The next generation of apps and agents will require trust built in. That’s why we’re building EigenCloud, a verifiable cloud for the agentic economy.

Tokenized assets aren't just static price trackers. They depend on real-time computation: NAV calculations, waterfall logic, compliance filters. Without verifiable computation, institutions are forced to either simplify the asset or reintroduce centralized trust. Read more 👇

Had a chance to give a first pass through this paper. This paper is really great. I don't think I have seen any other post-AGI economics paper that sharply defines the consequence of increasing automation and continuing relevance of human intelligence as the gap between cost-to-automate and cost-to-verify. There are two things that stood out to me. First one is this cycle: > when experts supervise agents, they produce corrections, labels, and "how-to" knowledge. That gets captured as training data / tooling which reduces the cost to automate. > as automation gets cheaper, firms replace humans with agents. That pushes need for humans down, especially for entry-level roles. > with less accumulated experience​, society has fewer capable verifiers and verification gets more expensive. > now you get the gap between cost-to-automate and cost-to-verify increasing, meaning agents can operate in more domains than humans can reliably audit (blind spots). > in those blind spots, humans can’t provide tight corrective feedback. Agents keep optimizing measurable proxies, and behavior can drift away from true human intent. Second one that stood out to me is the concept of "liability-as-a-service". Basically it is saying that the deployment of any agent by a vendor comes with a responsibility attached that if the agent causes harm / losses / legal trouble in defined ways, the vendor pays (up to a limit). So the customer is buying an indemnified outcome, not just software. This is basically forcing companies to internalize the risk from agents not being aligned with the user. I like this concept if liability-as-a-service a lot because the whole thesis of our @eigencloud platform was based on slashing and insurance. Any service deployed on @eigencloud is meant to be staked and if the service harms the user, the stake behind the service gets slashed and compensates the user based. on user's pre-purchase. I will highly recommend everyone to read this paper!

Eager to read: "as measurable execution commoditizes toward the marginal cost of compute, rents migrate to what remains scarce—verification-grade ground truth, cryptographic provenance, and liability underwriting (the ability to insure outcomes rather than merely generate them)."

8004 keeps expanding - tens of thousands of live agents, feedback signals and active users across 18 integrated chains. Dive into the latest stats on @8004_scan. The numbers speak for themselves 📈🤖

my cold email to @chamath

We closed a $4.5M seed round! Led by: @Maven11Capital & @FactionVC Also in: @cbventures, @signature_vc, @verisventures, @trgcapi, @dlabvc Building THE credit infrastructure for the stablecoin economy. $180M+ volume. Zero defaults. Let's go 🧵

@prodialabs ,one of the world's fastest image generators, uses Cloudflare R2 for AI model storage & transfer—calling it the "fastest and most economical solution". See how they built with Cloudflare 👇

It affords us the highest honour to proclaim the commencement of the Intmax mainnet. With the utmost respect and warmest salutations, we bid you welcome to Intmax. 1/8

The future of credit is onchain.

Huge news Talus Community! 🚨 Talus Devnet is officially live. 📄 Nexus Whitepaper is out. Today, we take a major step toward the Autonomous AI Economy. Nexus is our framework for building fully onchain Talus AI agents, and it is now live on our Devnet 🧵👇

Awesome collaboration between @flock_io and @Alibaba_Qwen . First of a kind partnership between a deAI project and top LLMs, for real! Excited to see what this brings...

The numbers speak for themselves — @Humanityprot is showing what real scaling looks like. 7.6M txs in 24h on testnet and scaling like a beast. Infra that actually performs — built by Gateway, powered by @0xPolygon CDK. That’s the stuff we need.

Announcing the biggest SQT rewards campaign in SubQuery History 💎 The 60 MILLION SQT Consumer Rewards Program is designed to boost projects using the SubQuery Network as their primary data API. Consumers can earn up to 900% of their query spending in SQT rewards. All you need to do is host and query your indexer on the SubQuery Network 💡 Whether you’re a native SubQuery project or a migrating Subgraph, the rewards are yours to earn. Start indexing today to secure a high rank once the programme starts! Why? Because builders deserve to be rewarded. Learn more ⤵️ blog.subquery.network/60-mil…

Welcome to Hemi mainnet! Today’s milestone comes with $440M in TVL (incl. $285M on our staking platform), and more than active 50 ecosystem partners! One Network, Powered by Bitcoin & Ethereum. Learn more. ⤵️ hemi.xyz/blog/hemi-launches-…

The Beras like RedStone. With PoL live now, RedStone is the oracle of choice for the bears. RedStone feeds secure the top protocols on BeraChain and support native assets like $BERA and $HONEY.