BlockWave🪽

A company that does not build a single AI model just hit $17.5 billion,, and it moves more AI traffic every day than Google or OpenAI👀 @FireworksAI_HQ Fireworks closed $1.5 billion this week. Nine months ago they were worth $4 billion,Now they are worth 17. The founder is @lqiao , the woman who built PyTorch at Meta before walking out and starting this in 2022. Here are the numbers that got them here :: --> $1 billion in yearly revenue, five times more than a year ago --> 40 trillion AI tokens moving through their system every single day --> Google handles about 27 trillion a day, OpenAI about 22 trillion, Fireworks does more than both --> 95% of that traffic is companies running their own custom models, not the generic ones Why every big enterprise is showing up here : )- Lin Qiao says it costs 5 to 10 times less than paying the top AI providers )- Companies tune their own models on their own data, and the model stays theirs )- Uber, Shopify, MongoDB, Cursor, and Harvey are already running real work on it )- The Ontario Teachers Pension Plan wrote a check too, and pension money does not chase hype For two years the loud story in AI was who shipped the biggest model. The quiet story was where the actual money was going,, Not to the models, to the layer is currently running them.

i've been testing @TermMaxFi v2 lately, and having orders across different chains in one place removes more friction than i expected what feels different is how fixed rate execution now sits beside limit orders and yield strategies without constantly pushing me between separate workflows the idle order update also matters in practice because capital waiting for a lending match can keep earning instead of doing absolutely nothing after watching the protocol expand its markets and collateral options, the product feels increasingly designed around keeping capital productive between decisions that is the part i appreciate versus typical lending flows, where moving from waiting to lending to leverage often means rebuilding the position worth spending some time inside the new flow and seeing whether the same capital efficiency stands out to you this post is intended for informational and educational purposes only and reflects personal views, not financial, investment, or promotional advice

GM builders, The hardest part of post quantum security may be teaching an app the difference between bad input and bad signature. Quip networks updated Rust hash signature library packages SHRINCS for Node and browsers through WASM. But the interesting detail is its failure contract: malformed hex throws a typed error wrong length data throws another a properly formatted but invalid signature simply returns false That distinction is not cosmetic. Wallets and apps need to know whether to reject a transaction, fix serialization, alert the user, or flag an attack. When every failure looks identical, secure code becomes guesswork. @quipnetwork is bringing post quantum cryptography closer to normal developer tooling, where adoption depends as much on predictable APIs as strong mathematics. Quantum safe apps will not be built by cryptographers alone. They will be built by developers who understand exactly why verification failed. Which matters more for adoption: stronger primitives or safer interfaces?

Weekend viewing: what "harvest now, decrypt later" means for your privacy, and how quantum computing is reshaping digital security. @rcarback, from his talk at last month's @zcash Blockspäti in Berlin 🇩🇪

x.com/i/article/2064…