Ash ‎Ashmawy

This is the kind of news that builds confidence. Massachusetts’ proposed $305M investment to support tech economy including AI expansion shows the state is serious about leading the next era of Physical AI. With top-notch universities, talent, and our tech ecosystem built for innovation, Massachusetts clearly has what it takes to win the race. The direction is right. Now there is only one mantra: scale. Excited for what this means for builders and for companies like CYVL working to turn innovation into impact for the real-world. #PhysicalAI #Massachusetts #AI #InnovationEconomy #CYVL #Scale

61

In 2018, two Wall Street technologists walked away from a planned crypto #hedgefund to build the missing engine room for institutional trading. That's the origin of Talos, founded by Anton Katz and Ethan Feldman. Both had cut their teeth at Broadway Technology, building low-latency systems that powered global markets. Katz went on to lead #tradingtechnology at AQR Capital Management, while Feldman sharpened execution engines that defined market structure. Their conclusion was blunt: #cryptoinfrastructure wasn't built for institutions. Their decision: build it themselves. Talos is not a venue, not an exchange. It is the connective tissue that lets hedge funds, #banks, and #assetmanagers trade digital assets with the precision they expect from equities or FX. The unified platform connects via FIX, REST, and WebSocket to 60+ liquidity venues, turning fragmented markets into a single entry point. Multi-dealer RFQ, smart order #routing, and customizable algos bring the tools institutions know. Under the hood, Talos runs modular microservices with Aeron messaging and Simple Binary Encoding, delivering sub-millisecond latency and zero-downtime upgrades. The scale is real. As of Q2 2025, Talos has powered $493B in cumulative volume across 5,242 symbols. It integrates 84 liquidity providers, serving clients in 31 countries. Adoption since 2020 is up fivefold. That's velocity. Capital followed. In May 2021, Talos raised a $40M Series A led by Andreessen Horowitz, with PayPal Ventures, Fidelity Investments, Galaxy Digital, Elefund, Illuminate Financial, and Steadfast. A year later, the Series B brought $105M at a $1.25B valuation, with Citi Ventures and BNY Mellon Ventures joining the syndicate. In 2023, Talos acquired D3X Systems (Acquired by Talos) for portfolio tools. By July 2025, it acquired Coin Metrics for $100M, locking down on-chain and historical data to deepen its moat. Culture matters. With offices in NY, London, SG, and Cyprus, Talos is global, diverse, and anchored by #reliability, #security, #innovation, and client obsession. Compliance is hardened with #SOC2Type1 and #Type2 audits and Big Four validation. Teams blend engineering precision, quant thinking, and client-first execution in agile loops for markets that never sleep. What separates Talos isn't just speed or integrations, it's trust. Institutions don't gamble with infrastructure. They align with partners who scale across cycles and deliver when volatility spikes. Talos has become that partner, embedding itself as the institutional standard for #digitalasset trading. Talos is hiring across #engineering, #product, #sales, and #operations. Link in comments 👇 #Startups #StartupSpotlight #StartupJourney #StartupGrowth #FounderStory #Crypto #Blockchain #Web3 #DeepTech #DeFi #FinTech #TradingTech #Infrastructure #Compliance #Technology #Innovation #TechEcosystem #StartupEcosystem #Hiring #TechHiring If software engineering peace of mind is what you crave, Vention is your zen.

1

1 Comment

🚨 A rare sight: 10+ vaults with paused deposits on the official Morpho UI following the Resolv Labs USR depeg from yesterday. Folks affected through the ecosystem through second-order effects, e.g. buying cheap USR and using it as collateral to borrow USDC on lending markets with oracles harcoding USR at $1. Painful reminder 0 risk does not exist (Resolv had 15+ audits. The code worked as designed during the attack, it was opsec failure that allowed attacker to mint 80 million USR on ~$200k collateral). Vault-native cover 🔜

13

3 Comments

🚨 𝗡𝘃𝗶𝗱𝗶𝗮 𝗜𝘀 𝟱𝗫 𝗠𝗼𝗿𝗲 𝗘𝘅𝗽𝗲𝗻𝘀𝗶𝘃𝗲 𝗣𝗲𝗿 𝗧𝗼𝗸𝗲𝗻 𝗧𝗵𝗮𝗻 𝗚𝗿𝗼𝗾. The whole AI cost conversation just changed overnight. 🔍 𝗧𝗛𝗘 𝗗𝗘𝗘𝗣 𝗗𝗜𝗩𝗘 ⭕ Groq charges 5 to 10 cents per million tokens while Nvidia's B100, B200, and B300 run 25 cents, a 5X premium on the same output. ⭕ Groq delivers up to 800 tokens per second versus Nvidia's 450, nearly double the throughput at a fraction of the price. ⭕ The AI spend model is flipping from dollars per GPU hour to dollars per million tokens, and procurement teams are catching on fast. 🤔 𝗪𝗛𝗬 𝗜𝗧 𝗠𝗔𝗧𝗧𝗘𝗥𝗦 Your inference budget is about to get benchmarked against cost per million tokens, not instance hours. Teams still anchoring forecasts to GPU line items will look asleep at the wheel when Finance asks why Groq wasn't in the evaluation. 🚀 𝗬𝗢𝗨𝗥 𝗡𝗘𝗫𝗧 𝗦𝗧𝗘𝗣𝗦 Pull your current inference bill and calculate your real cost per million tokens this week. Repost this for the procurement lead on your team who still thinks GPU hours is the only metric that matters. Read further: https://vist.ly/42276 #CostPerToken #InferenceEconomics #AIChips #finops #ai

1

The "Dollar per Token" model is confusing, unpredictable, and increasingly decoupled from actual performance. 📉 I ran a series of tests to show how wildly your "hard-won dollars" fluctuate when translated into tokens. The results were all over the place. Note: this article is not exhaustive and there are likely better cost analysis/benchmark articles out there, but it’s just meant to show how token pricing is very confusing and not at all correlated to a model’s performance on a benchmark. Given prices change frequently, I am using OpenRouter prices as of late April 2026. Since sharing is caring, I’ve made the scripts public so you can run your own benchmarks (or improve them): https://lnkd.in/dVbB-mpu https://lnkd.in/d5q2Jhrv

16

3 Comments

Everyone's building Agent Pay. Most of them are building half of it. x402 from Cloudflare and Coinbase handles the HTTP payment standard. Circle Nanopay handles micropayment denomination. Visa Intelligent Commerce handles network acceptance. These are serious pieces of infrastructure — and they all solve the on-chain leg. None of them address what happens when the agent tries to pay at a real-world merchant. 90% of commerce still runs on card acceptance. An AI agent with a stablecoin wallet and no card programme cannot pay at most merchants, book most hotels, or settle most supplier invoices. The agentic payment stack has three non-negotiable layers: a programmable on-chain wallet, card issuance for merchant acceptance, and a compliant KYC/AML framework for a non-human cardholder. Solving one or two of these is progress. It is not a product. Wirex operates all three through a single program. Not because we anticipated the "Agent Pay" narrative — because the underlying requirements are the same ones we have been solving for corporate, prepaid, and crypto-linked cardholders for a decade. The category just found its name. The issuer-side problem — underwriting and ongoing compliance for an autonomous agent as the primary cardholder — is the piece nobody in the current conversation is addressing publicly. That's where the hard work actually sits. Who's your issuer? #Wirex #payments #stablecoins #crypto #BaaS

35

10 Comments

Blockchain matters far beyond crypto. It is the infrastructure layer behind how information, identity, ownership, and value move digitally. This clip is from my latest article breaking down what blockchain is and why it matters.

10

1 Comment

Too often, Forward Deployed Engineers (FDEs) are mistaken for professional services. They’re not. Marc Benioff highlights Palantir’s FDE model as a playbook for how to embed technical builders directly with customers, not as vendors, but as co-architects. FDEs accelerate product-market fit, and create a live feedback loop between engineering and customer impact. In the AI era, FDEs are not optional. They are the connective tissue between vision and outcomes. https://lnkd.in/gfi5WxQh #FDE #AI #Disruption

5

DealWire.TECH Commentary 20.04.2026: $6.6B left DeFi in 24 hours — and nothing was “hacked.” Everyone is calling it panic. It isn’t. It’s recognition. A single compromised validator in a bridge (via LayerZero Labs) enabled an attacker to drain ~$292M from Kelp. Those assets were then routed into Aave Labs, used as collateral, and turned into real liquidity. No smart contract exploit. No broken logic. Everything worked exactly as designed. And that’s precisely the issue. Because what lenders on Aave suddenly realized is this: They weren’t just lending against collateral. They were underwriting the entire upstream security stack behind that collateral — bridges, validators, signatures — without ever pricing that risk. The market corrected instantly: Billions exited not just Aave, but across protocols. Unrelated platforms saw liquidity flight. Different chains, same reaction. This is not contagion. It’s systemic coupling. Now extend this one step further. We are entering a world where AI agents will: - Move assets autonomously - Optimize yield across chains and platforms Interact with APIs, wallets, exchanges, and lending protocols in real time In other words: AI will sit on top of already composable, already fragile financial stacks. And AI does not “hesitate.” It scales decisions instantly. If a human panic causes $6.6B to move in 24 hours, an agent-driven system can do the same in minutes. The real risk is no longer just hacking. It’s delegated trust. Every additional layer—whether it’s: a bridge, an API, an identity provider, or an AI agent acting on your behalf …becomes part of your attack surface. But most users, and even institutions, still think in silos: “My account is secure.” “My platform is secure.” That model is obsolete. Security today is not about code correctness. It’s about system integrity across layers you don’t control. And increasingly: Across decisions you don’t even make yourself. TradFi often looks slow. But this week, it was reminded why: Friction is sometimes the only thing standing between stability and cascade failure. As AI agents begin to manage capital, identities, and operations, the question is no longer: “Can this system be hacked?” It’s: “Do we actually understand the full chain of trust we are delegating to?” Most don’t. And markets are starting to price that truth. #DealWire #CyberSecurity #AI #DeFi #RiskManagement #FinTech #AIagents #Blockchain #Trust #DigitalInfrastructure

4

The biggest risk in an exit is the earn-out. Most founders sell their firm only to realize the secondary payout depends on performance metrics they no longer control. If the revenue dips because you stopped supervising every case or client relationship, your exit value evaporates. This is where systems architecture becomes your safety net. At HHCE, we help law firm owners transition from supervising people to commanding systems. As outlined in The AI General, the objective is to build an empire engine where operational functionality remains in your absence. When you replace manual oversight with autonomous operational systems, the business results become predictable. Predictability is what protects your earn-out. We build the architecture that ensures the firm scales even after you hand over the keys. Your exit should be a clean transition, not a multi year gamble on your own exhaustion. Build a platform that is inevitable, not personality dependent. Let us audit your current operational structure and secure your legacy. Book a briefing: https://lnkd.in/eHUaQf4w

Amazon.com's AWS revenue growth and spending are accelerating, yet unlike the largest AI hyperscaler peers its current policy -- which includes no capital allocation toward shareholder returns (dividends or share buybacks) -- limits the company's need to raise debt to fund expansion. Exiting 3Q with $94 billion of cash and plans to spend $125 billion in capex this year and more in 2026, its ability to maintain cloud leadership persists while generating positive excess bond returns.

1

Robinhood is about to overtake Coinbase in market cap. Helps to be dominating in crypto and stocks. Coinbase needs to get in the equity markets and tokenize equity. Benefits of Tokenizing Equity: • Faster settlement (seeeconds not days) • Lower fees (no middlemen like brokers or clearinghouses) • Fractional ownership (own 0.000001 shares) • Global access (24/7 markets) • Programmable (e.g., automatic dividends, governance via smart contracts) Get going.

17

1 Comment

The Headless Bank: Sigma BankAI's I'M IN AI OS Pivot The provided text outlines a dramatic strategic shift by Sigma BankAI known as the "I'M IN AI OS" pivot, which moves away from traditional web-based interfaces toward a "headless" infrastructure. This transformation prioritizes autonomous AI agents over human users by exposing all banking functions through APIs, command-line tools, and the Model Context Protocol. By rebuilding its ecosystem as a programmable nervous system, the company allows external agents to execute complex financial workflows, such as tokenomics management and remittance routing, without needing a graphical browser. The narrative highlights a broader industry trend where system transparency and agentic execution replace manual navigation, effectively turning the banking platform into an invisible foundation for digital labor. Furthermore, the shift introduces consumption-based pricing and new programming languages to ensure that AI operations remain deterministic and secure at scale. Overall, the sources depict a future where context and data serve as the core value, rendering the legacy web browser obsolete in favor of integrated, intelligent automation.

1

2 Comments

2% NAV slip. Almost nobody saw it. Midas mF-ONE, a $151M tokenized Fasanara credit pool dropped $3M. Private credit in DeFi pretends to be stable. Until it isn’t. NAV cuts don’t announce themselves. Opaque underlying exposures mean retail and institutional holders are often the last to know. mF-ONE didn’t “break.” It just simply repriced from $1.06 to $1.04. Tiny on a chart, annoying if you thought yields were predictable.

33

4 Comments

Just published a new white paper on a structural dynamic that I think investors should pay attention to. Stablecoins have become one of the largest marginal buyers of T-bills, but the way they’re regulated creates a convexity trap. When redemptions hit, the forced selling of bills can push yields up 2–3× more than inflows push them down. As the market moves toward trillion-dollar scale, this becomes a real signal for macro, liquidity, and venture theses. This is a rare case where crypto market structure directly influences sovereign debt pricing. For funds looking at stablecoin infrastructure, tokenized cash, or macro-sensitive fintech, this dynamic will matter. Interested to hear reactions from investors across digital assets, macro, and new financial infrastructure. #VentureCapital #Fintech #DigitalDollar #Stablecoins #Macro #Liquidity #Treasuries #Tokenization #InvestmentThesis

1

Robinhood recently launched a new AI feature called Cortex Digests. It’s meant to give investors a plain-language answer to the hardest question in finance. The digest scans news, analyst reports, and data, then spits out a one-liner like: “Shares rose on strong earnings, analyst upgrades, and an AI partnership.” For retail investors, that’s powerful. No jargon, no hours of research, just quick clarity. But the system leans heavily on a small set of outlets, one of the main ones being Benzinga, which runs paid and sponsored posts. That means a contributor or advertiser could effectively buy their way into Robinhood’s AI digest. Seed a narrative → the model ingests it → the digest repeats it as “market context” → traders act on it. You can already see this dynamic in action with CRDO, where Cortex pointed to analyst upgrades, AI deals, and revenue growth as drivers of a rally. Those were real but if the sources tilt, so does the story. And when millions of retail users see the same digest, the effect multiplies. AI in finance can empower small investors but only if it’s transparent about sources, resists manipulation, and is safeguarded against the same risks we see elsewhere in generative AI: hallucination, data poisoning, and prompt injection. Otherwise, it risks becoming a financial echo chamber and one that doesn’t just amplify markets, but actively shapes them based on whoever learns to game the system first. Markets move on trust. And if AI is going to guide investors, it needs guardrails as strong as the insights it provides.

15

2 Comments

Lido Finance's V3 & Kiln: Expanding Institutional #Ethereum Staking with stVaults 🚀 Kiln participates in Lido V3 as an approved node operator, operating enterprise-grade validators that directly back stVault deployments. ETH associated with these vaults is staked through Kiln’s institutional validator infrastructure. stETH minting is optional, while all validator monitoring, performance management, and operational overhead are handled by Kiln. The setup benefits from: ✅ Vault-level asset isolation with transparent, on-chain attribution ✅ Kiln’s SOC 2 Type II–compliant validator infrastructure ✅ Lido V3’s audited stVault smart contract framework This combination ensures institutions can access liquid staking functionality with strong guarantees around security, uptime, and performance differentiation. Learn More: https://lnkd.in/ds_CJF9e

91

2 Comments

One product, two markets.... Stablecoins are being pulled in two directions as institutional teams want settlement speed and liquidity control and retail users still need trust, utility, and simple on and off ramps. When one product tries to serve both, the roadmap gets confused. You either build for compliance and treasury flows, or you build for consumer habit change. They are not the same job, and they do not need the same messaging or commercial model. Many teams are attempting to launch both at once, which usually means neither is great. The institutional roadmap gets slowed by consumer UX debates, and the retail roadmap gets slowed by governance requirements. I do not see enough clarity on this split. If you had to choose one market to win first, which would you pick, and why? #stablecoins #payments #fintech

3