Market speculation is NOT about markets and prices and trading hours and margins and charts and government reports Have any of you figured this out yet or is the con still sending you down the "road of easy answers"?? Markets are all about redistributing wealth from the many to the few (of course, not in every market every year) The real game is played in the upstairs private poker room (commercials, reportable specs) You and I (5 to 20 or so futures contracts a trade) do not even count. We are peons. Ants viewed from space That the avg spec has not figures this out yet is the reason that fewer than 3 in 1,000 who try make it People whine about their losers. If is always a story of coulda, woulda, shoulda or the latest gimmick in charting or a bad government report. Anyone who believes these fables deserve to lose their money. Seriously, that is my view. Every once in while I see on X a post by someone bragging about the hypothetical profits they made from the "sim trade for a fee" scam. The sooner a retail speculator realizes that trading stocks and commodities is just a huge computer game the better. I view long term holds in quality companies as an investment. Being married to my wife of 58 years is an investment. Not having debt is an investment of sorts. Anyway, this is how I think about it Peter Brandt Founder and CEO, Factor LLC and The Factor Report [As a personal note from me regardless of what you do with your money, surrendering my heart and soul to Christ was the ultimate investment whose dividends will never end.]

FOUNDERS FRIDAYS EP. 13: @AdrianBrz_ from @ridemarkets 1️⃣ What makes you most bullish about @Solana today? The cost structure, and what it does to behaviour. Look at tokenized equities: Solana holds under a quarter of the ~$3B of tokenized stock supply on-chain - BNB Chain and Ethereum both hold more. But almost all the actual trading happens on Solana. Issuance follows regulation and distribution deals; trading follows execution cost. When a transaction costs a fraction of a cent, people trade twenty times a day with $5 at stake, and that behaviour simply doesn't exist elsewhere. Every chain wants to be where assets get minted. I'd rather build where they get used. 2️⃣ Where are the biggest hidden opportunities for new builders? Anything to do with how capital actually gets allocated. I've spent the last five years on both sides of it building DAO governance infrastructure and working on DeFi exchanges and the same pattern shows up everywhere: we have endless ways to trade and almost nothing that decides who gets to deploy capital and on what terms. Most DAO treasuries just sit there while a few people on a multisig argue about a proposal. The same gap is now opening in tokenized equities the assets are arriving fast, but the things you'd want around them (credit, structured products, portfolio tooling, anything that treats a tokenized stock as collateral rather than something to flip) barely exist yet. Governance tooling and treasury ops are the same story: unglamorous next to a new DEX, and far less crowded. 3️⃣ What’s your number one piece of advice for someone starting from zero? Ship something small in public before you feel ready, then keep cutting until one thing works properly. Do it where people can see it post the ugly early version, contribute to open source, fix someone else's issue. Distribution compounds slower than code but it lasts longer, and most of the opportunities I've had came from work that was public before it was good. And the thing nobody has really priced in yet: since January I write almost no code by hand I write the spec and review what the model produces. If you're starting today, the skill worth building is judgment, knowing what to ask for and what to reject, not typing speed. That's a real advantage for beginners, because the thing you'd have spent three years grinding is no longer the bottleneck.

🗡️ What DeFi Shipped 🚢 (Sept 23): Circle launched its own L1, and Aave and Compound both built doors for institutional money. Here are 14 big updates in under a minute: • @Arc, the L1 from @circle, went live. TVL (which could be rented) looks solid, but the reception across CT was poor. • @aave proposed custodied collateral lending with @Anchorage holding the collateral. Institutions could borrow stablecoins against Bitcoin that never leaves regulated custody. • @Compound_xyz opened its Institutional Market for USDC borrowing, with LTVs up to 87% on ETH, 85% on wstETH, and 81% on wBTC and cbBTC. • @pendle_fi went live on Robinhood Chain with yield markets on tokenized stock dividends, starting with NVDA maturing October 15 and PFE maturing December 10. • @Uniswap shipped StablePair Hook, a v4 hook that sets LP fees based on how far a pool has drifted from its reference rate. Live on two Ethereum pools, USDC/USDT and USDC/USDG. • @base published the spec for Validity Transactions, which the chain includes only when onchain conditions match, gated on balance, storage, block number, or Flashblock index. • @MetaMask now auto-reverts a transaction when real execution doesn't match what the wallet simulated. It targets "red pill" attacks, where a contract behaves one way in simulation and another onchain. • @Pumpfun replaced Cashback Mode with Holder Rewards. Holder Reward tokens now pay out just for holding, with higher ceilings the longer you hold. • @LidoFinance and @Stakely_io opened a public ETH staking vault built on stVaults, pairing ETH staking with the EarnETH DeFi strategy. • @RenzoAI launched Renzo Basis, an automated delta-neutral basis trade on @HyperliquidX supporting BTC and HYPE at launch. • @zama scaled up confidential DeFi with 16 yield vaults across five institutional curators, covering USDC, USDT, WBTC, AUSD, and tGBP, all deployed on @Morpho. • @0xMiden, the privacy-focused chain incubated by @0xPolygon, shipped testnet v0.16, the last major release before mainnet. • @DeriveXYZ proposed V3, a zkVM exchange settling on Ethereum L1. The matching engine and sequencer state transitions get proved in a zkVM, with user funds custodied in L1 contracts. • @verantaxyz, the leveraged-trading protocol on @base, rebranded from Avantis and dropped its invite-code gate. I’m fine with institutions keeping their custodians and compliance teams bc that’s probably what it takes to get them comfortable using DeFi. I just don’t want that to become the only way to use it. I’d still like to borrow against my holdings at 2 a.m. without asking someone for permission.

I want Ethereum and the EVM to win. A strong EVM flows back to everything built on it. Every decision in 8130 considers the goals of the various EVM chains, the wallet, the app and the user. The goals of each are different but they don't have to be in conflict. One simple test for a native AA standard: does it serve everyone's needs? Any standard can be enshrined on one chain. Only one that serves every constituency (the chain, the wallet, the app, the user) gets adopted everywhere. So that's what 8130 was designed for, from the ground up. All chains want performance, but differ on which features they need or want to support. Features and performance can typically compete for the same resources, so 8130 put a lot of work into decoupling pieces that are normally conflated and allowing each chain to make the right choice for itself. The foundation of 8130 is performant, feature rich and easily extensible. That's why one standard serves high-performance L2s and alt-L1s (with goals for scale, short block times and custom feature sets) and extends to everything L1 needs: privacy flows, maximum decentralization and censorship resistance. And of course a post-quantum path for all. Wallets want great UX with minimal overhead. Familiar systems and not a custom implementation per chain (though that's the way it's been going). Ideally one standard that works everywhere and in the same way. 8130 treats portability as paramount. Build once and it's always one transaction to deploy, sync, and act on a new chain with the same address everywhere. Apps want the same thing one layer up. Build for one account model without reasoning about wallet compatibility. Standardized session keys, subscriptions, and permission grants remove an entire class of problems. Native AA means one-click passkey onboarding and sponsored transactions out of the box. The end user gets what all of this was for: accounts that work everywhere, gas they never think about, one click instead of five signatures, keys they can rotate and recover without moving funds. Apps that just work. Now do the same test for the alternative AA proposal. With frames, chains must choose between features and performance. One example is 8141's default-account fast path as it doesn't natively provide passkeys, session keys, sub-accounts, or key rotation. Those require custom authorization code, which requires tracing and forfeits the fast path's direct evaluation. Chains that can't afford that, or other inherent tradeoffs in the spec, may not adopt it, or if they do, they may have to adjust their goals. Wallet fragmentation persists, apps keep building for multiple targets, and overall UX suffers. L1 enshrining one standard that can't satisfy the needs of the wider EVM ecosystem results in divergence. I believe this is avoidable. 8130 can be a unifying account standard for the EVM, and we can continue toward the world-class UX that crypto deserves.

# Practices for Embedding AI Agents in Software # Sync Facade over Async Core 🎯 The Hook Choosing between "always sync" and "always async" is a false dilemma. What if your API could return instantly when fast, and gracefully degrade when slow? 🔥 The Problem Agent processing latency follows a bimodal distribution. Cache hits and lightweight tasks return in milliseconds, but complex reasoning or tool chains stretch to tens of seconds. Always-sync leads to timeouts and connection exhaustion. Always-async forces polling even for sub-second responses. 💡 The Pattern The Sync Facade always processes internally via an async pipeline. The outward-facing API waits up to a configurable threshold: if the job finishes in time, it returns a 200 with the result; if not, it returns a 202 with a job ID for async retrieval. Clients hit a single unified endpoint without worrying about latency bimodality. The threshold is tuned adaptively based on observed P95/P99 latency trends, not hardcoded. ✅ When to Use Use when: - Latency distribution is bimodal (mix of fast and slow completions) - Existing clients expect a synchronous API contract - Latency requirements vary per request (chat UI vs. batch) Don't use when: - Processing always finishes in a few seconds (use Sync Edge) - Processing always exceeds 30s (use Durable Async from the start) ⚠️ Pitfalls - Do not hardcode the sync-wait threshold. Observe P95/P99 trends via tracing and adjust adaptively - Clients often miss 202 handling. Explicitly define the 202 response schema in your OpenAPI spec to prevent SDK generation gaps - Distinguish worker crashes from simple timeouts during the sync wait. On worker failure, escalate to 202 immediately rather than waiting for the threshold 🔧 Implementation Approach - All requests are internally processed via an async queue. The facade layer awaits the result up to a configurable sync-wait threshold, returning 200 with the result on success or 202 with a job_id on timeout - Tune the sync-wait threshold adaptively by observing P95/P99 latency trends via tracing. Starting points are roughly 5-10s for web APIs and 30s for internal RPCs - Use SSE or WebSocket for progress notifications after async escalation, with polling (3-5s interval) as a fallback for clients that cannot maintain persistent connections - Keep the facade layer as a thin adapter with no business logic. The async core reuses the Durable Async Agent checkpoint and resume machinery directly - Explicitly define the 202 response schema in the OpenAPI spec so that generated client SDKs correctly handle job ID retrieval and result polling #AIAgents# #SoftwareArchitecture#

Spec sheets sell FLOPS. Decode spends bytes. Single-user BF16 decode can sit at roughly 1 FLOP per byte. On an H100 SXM, that puts its theoretical ceiling near 0.3% of peak math throughput.* Batching moves the point right: more tokens share each weight read. FP8 has a twist. Halving weight bytes moves the point right, but native FP8 math doubles the compute roof too. The ridge moves with you. Decode gets faster, yet the batch needed to reach the roof stays about the same. Attention limits the gains. Each user brings a KV cache. Longer contexts mean more bytes to read, so sharing weights buys less. Smaller caches and speculative decoding can help. *Teaching estimate for weight-dominated, batch-1 decode, not measured GPU utilization.

Spec Decoding (DSpark) finally works with Pipeline Parallelism in vLLM!! 🔥 A lot of the GPU poor/proletarian have been asking for this feature for a while now, but it took the recent Kimi K3's massive 2.8T parameters affecting the GPU middle class (B200) for it to be implemented! Before this change, GPU poors that needed to use pipeline parallelism, as their weights didn't fully fit on 1 server, would not be able to take advantage of DSpark. Shoutout to yongqinwang-cmd & Inferact for implementing it in vLLM!

The Honda Passport Baja Spec is a new state-of-the-art trophy truck that features a slew of advanced tech. Along with an 800-volt hybrid system, it has a 1300-plus-hp twin-turbocharged 2.4-liter V-6 based on the engine in Acura’s IMSA race car. The Passport Baja Spec’s first competition will be the 2027 SCORE Baja 500 in June. More here ➡️