Alva

Introducing Alva’s FinTwit Alpha Leaderboard. We burned $100K tokens backtesting 3000+ FinTwit accounts and ranked who actually makes money. $1M in reward is going to the top accounts across all our leaderboards.

@tradergokux how therapeutic! amazing view.

My friend from the ocean.

also some of our 0.02: Feel like Burry can hold all of these at once: 1. AI market looks bubble-like That is a broad market valuation/risk view. It does not mean “crash next month” or “short every cyclical stock today" 2. Prediction markets threaten sportsbooks This is more of a industry-structure view. If prediction markets are curbed by regulation, $DKNG/$FLUT benefit because their regulated sportsbook moat gets protected. 3/ DKNG/FLUT can still be vulnerable in a recession True, but that does not kill the trade if the catalyst is regulation, consolidation, or prediction-market restriction before a macro bust. So re “If the AI bubble bursts, gambling companies get hit hardest, therefore DKNG/FLUT must be short-term and inconsistent", that only works if you assume Burry is making a single all-in macro timing call. Also, 13F positions are pretty delayed. He could have hedges, could already be out, or could be sizing it as a specific catalyst trade (Burry has done plenty of tactical weird trades).

I'm so confused with Michael Burry. > Says that the market today resembles the dotcom bubble because of AI. > If the market bubble bursts, the global economy would enter into a violent collapse. > As a result, gambling companies would be one of the main sectors to be impacted hardest as a second-order effect, just like prior recessions. Therefore, his $DKNG and $FLUT bets have to be very short-term plays since his thesis timelines don't match up at all. Which seems to go against his whole investing philosophy of taking big, bold, long-term bets.

re the screenshot, feel like Burry’s likely thesis is more like prediction markets are the real threat, regulators will curb them, and that protects licensed sportsbooks. So if Kalshi/Polymarket-style event markets can offer sports-like contracts under a lighter federal/CFTC framework, they could attack sportsbooks with lower taxes, different licensing, and potentially better pricing. That's a threat to DKNG and FLUT. If regulators clamp down, the existing state-licensed sportsbook moat gets stronger. So the bet (per screenshot) is probably - Long DKNG/FLUT: incumbents benefit if prediction markets get boxed in. - Regulatory catalyst: courts/agencies/states decide what prediction markets can offer. - Consolidation thesis: FanDuel/Flutter and DraftKings remain the scale winners. Not necessarily macro bullish. He can still think the AI market is frothy while owning a specific regulatory-moat trade.

@leto_bao 🙌 Appreciate the integration of Alva’s AI investing agent! Great to see Alva put to work here. If the community has any feedback, we’re always listening.

欢迎加入Discord群！目前已经有 22222 人，近期新增了很多 Agent 板块。 每日总结：群里消息太多没空看，看Agent总结，一键跳转到感兴趣的话题。 leto的x：我的X的解读。 异动雷达：市场最新消息快速获取。 alva投资助手：让 AI 从多渠道数据源帮你解读投资信息。 所有功能免费

interesting... If Iran-linked ships are still moving while others pause or reroute, it points to selective danger rather than a simple Strait-open-or-closed situation. That can still be disruptive because insurers, charterers, cargo owners etc have to price the chance that their ship is the wrong one in the wrong place. A formal closure is not required for Hormuz to become expensive.

Renewed fighting this week between Iran and the U.S. has rocked shipping traffic through the Strait of Hormuz—except for Tehran’s own vessels on.wsj.com/44nD5p8

@michaeljburry issuance and buyback totals can blur the point. The better & sharper datapoint would be net share count by sector (AI buildout, energy capex, and balance-sheet repair are very different cases than mature cash generators simply stopping repurchases)

Companies are issuing shares more, buying back less.

Check it out here: alva.ai/u/lake/playboo…

Here's another very cool and comprehensive humanoid robots live theme tracker we built, tracking its supply chain, including 83-name basket × 9 layers × 3 pillars, thesis, chokepoints, catalysts Subscribe to this playbook, and you will get live notifications when any catalysts/news/chokepoint change. Playbook link in comments.

This is interesting if the law suit is really about OpenAI’s hardware ambitions, not just the legal claim itself. Apple usually fights hard when it thinks a new product category could weaken its grip on the consumer device experience. would be careful taking the filing at face value before seeing the complaint

*APPLE SUES OPENAI IN FEDERAL COURT FOR THEFT OF TRADE SECRETS

wow thought they had a partnership (although that partnership was about bringing ChatGPT into Apple Intelligence) Apple can partner with OpenAI on the model layer and still fight them on the device layer ChatGPT inside Apple Intelligence is one thing. OpenAI trying to build consumer hardware with ex-Apple talent is a much more direct threat to Apple’s control of the interface

@unusual_whales interesting. curious to see the actual complaint here. “Trade secrets” could mean anything from narrow employee/IP claims to something much closer to how AI gets built into devices

BREAKING: Apple, $AAPL, has filed a lawsuit against OpenAI, alleging misappropriation of trade secrets.

Search any hot stock and quickly see what investors are actually watching about this stock. Check it out below. alva.ai/u/lake/playboo…

The worst part of stock research: You can spend 6 hours reading about a company on 30 different tabs and still not know what investors are watching for this stock & what this stock is actually trading on... That’s what this live playbook fixes. Search any hot stock and quickly see what the market is focused on right now: market narrative, what could move the stock, market focus over the past 4 quarters, competitors, and where the Street stands. Check it out! Playbook link in the comments.

@MarioNawfal Interesting analysis & great interview I think the broad concern is fair, but the framing is a bit too apocalyptic. Treasuries don’t need to “stop being safe” for higher yields or bad liquidity to cause real pain across assets

Technical Analyst and Market Strategist Michael Oliver says everyone is watching the Iran war, but the real crisis is already forming inside the U.S. bond market, and when it breaks, it could hit everything. For decades, investors have treated U.S. government bonds as the safest asset on Earth. Michael says that assumption is beginning to crack. He argues the real crisis isn't inflation, it isn't recession, it isn't even the Middle East. It's the growing possibility that confidence in government debt starts to break down. If that happens, the Federal Reserve will have to create even more money to support the bond market. And he believes investors are already starting to prepare for that shift by quietly moving into real assets: gold, oil, industrial commodities, and agriculture. Assets that can't simply be created with another round of monetary expansion. He also pointed to something that rarely gets discussed. The biggest bubble is the belief that government debt will always remain the world's safest investment. If that confidence disappears, the consequences won't stay inside the bond market; it will ripple through virtually every corner of the financial system. Most of the world is focused on the next missile strike on Iran, but he's watching the next Treasury auction. Because in his view, history won't remember the Iran war as the event that changed the markets. It'll remember it as the distraction that kept everyone looking in the wrong direction while the real crisis was gathering underneath their feet. @Oliver_MSA

@zerohedge A 13% move after a big US offering says people still want scarce AI memory exposure badly enough to look through dilution. But I personally would not treat the offering itself as automatically bullish though. It still has to fund capacity that earns great returns.

*SK HYNIX ADRS CLOSE 13% HIGHER AFTER $26.5 BILLION US OFFERING

@StockMKTNewz Looser UAE controls could help Nvidia if this turns into licenses/orders/and shipments

The 🇺🇸 loosened export controls on the United Arab Emirates 🇦🇪, making it easier to export Nvidia $NVDA AI chips, military equipment, commercial satellites and spacecraft rallies.ai/news/us-makes-…

@business The funding headline is helpful, but it is not the whole story. Mexico needs clean power for industrial load with nearshoring. Need to see whether these projects can actually get permitted and bring in private capital alongside Banobras.

Mexican finance officials are exploring the possibility of creating an umbrella-financing package for renewable-energy projects that could be supported by more than $4 billion from development bank Banobras. bloomberg.com/news/articles/…

@LukeGromen Useful chart! but the comparison to DoD spend can make this sound like a direct tradeoff with current weapons budgets. past wars leave long-tail fiscal costs. I wonder how much of the rise is benefit generosity, eligible population, healthcare inflation, or some mix

US Veterans' Benefits payments as a % of Total US Defense Department spend, 1959-present. Veterans' Benefits now a record 27% of DoD spend (and ~8% of total US Federal receipts), rising far faster than both DoD spend & Federal receipts.

yeah, HBM tightness is one of the few AI bottlenecks that can show up everywhere at once: GPU shipments, cloud capex timing, and memory margins. The question not figured out yet is who actually keeps the economics when supply is scarce: HBM makers? packaging capacity? GPU vendors? or the biggest buyers with locked-in contracts?

SK Hynix's CEO warned that the global memory chip shortage could continue beyond 2030 due to rapidly growing AI-driven demand. Demand for high-bandwidth memory (HBM) used in AI chips is expected to outpace supply for years. Expanding production is difficult because of the high cost, long construction times, and complex manufacturing process. The company believes the semiconductor industry must significantly increase investment to meet future AI-related demand.

yes, orchestration is not just choosing the best model for a prompt. Once agents touch real workflows, the layer has to preserve state, permissions, evals, cost limits, and failure handling across the chain. That starts to look more like operating infrastructure than a thin router.

Summary: I spent time trying to figure out this orchestration layer problem, can we design a multi model architecture in the long term. The more I dug in the more I understand that trying to build an abstracted layer is hard. As agentic activities increase and agent chaining and complex tasks get assigned to AI it will become harder to move between models. There is a reasonable probability that 75% of the enterprises will build their implementation of the solution to their core problem around one model "stack". Token price reduction by 90% is the solve and mobility between models from the same frontier lab! Evals, harnesses, cache memory are the moats and I don't see models providing simple abstraction to those. I know there are efforts to do this out there, the long term solve for orchestration if it works will need to be "Claude code" level of design genius. Here's a chat with Fable @HamzaFodderwala had. **Why abstraction looks easy.** Models are stateless — every API call is weights + a prompt assembled at runtime. Everything the model "knows" about you — memory, documents, history, tools — is injected into the context window by software outside the model. So in principle, all your state already lives outside the weights. The catch is what "state" includes. **Layer 1 — Data (fully portable).** Enterprise documents, tickets, logs. Retrieved via RAG: text is chunked, embedded, stored in a vector database (Pinecone, pgvector), and relevant pieces are fetched into the prompt per query. The embedding model is separate from the LLM, so this layer is genuinely model-agnostic. Already solved. **Layer 2 — Memory (portable in principle).** Systems like Mem0 and Zep sit between the app and the model: after each interaction they extract salient facts ("user prefers X"), store them as plain text, and inject the relevant ones into future prompts. Because the artifact is natural language, it reads into any model. Facts port. **Layer 3 — Orchestration/routing (works, but only for shallow tasks).** Gateways like OpenRouter and LiteLLM normalize API differences and route each request to the cheapest capable model. This is the fungibility layer being furiously built. It genuinely works for one-shot, verifiable tasks — classification, extraction, summarization — which conveniently are the tasks where cheap models suffice anyway. **Where it breaks — the non-portable state.** Four things stay behind when you switch: - **The harness.** Prompts, tool schemas, and guardrails are tuned to one model's quirks. An agent must get every step right, so reliability compounds: a model that's 98% reliable per step completes a 50-step task about a third of the time; at 90% per step, it almost never finishes. Swapping models costs you a few points per step — the difference between an agent that works and one that doesn't. - **The evals.** Swapping means re-testing everything and re-fixing every regression. The real switching cost isn't data migration — it's re-verification. Nobody has abstracted that. - **Procedural memory.** Facts port; skills don't. Cached successful workflows and learned workarounds are conditional on the model that produced them. - **Cache pricing.** Provider-specific, worth 75–90% of input costs on agentic workloads. Quiet lock-in. **The labs' angle.** They offer hosted memory, hosted file stores, caching, fine-tuning — every one pulls state from your side onto theirs. The labs will crack memory first, but as lock-in, not portability. Nobody standardizes their own exit door. MCP is the partial exception: it standardizes tool and data access across models, but doesn't touch harness tuning or evals. **Where 3P vendors fit.** Routers are thin-margin commodity plumbing; vector DBs and memory infra are real but small. The two structurally interesting positions: **eval platforms** (LangSmith, Braintrust) — since switching cost equals re-verification cost, whoever industrializes cross-model testing actually enables fungibility.