Adrian x Market

@MoatReview @marvel_updat3s The Punisher MAX run exists because the character only works without guardrails. PG-13 Frank is just a guy in a skull shirt who's really upset.

@marvel_updat3s [Frank Castle's entire identity is built on zero compromise, that's the character. "Fun workarounds" for the killing is exactly the kind of executive note that turns a great crossover concept into a sanitized version nobody asked for.

Tom Holland says Punisher will still be “authentic to the Frank Castle we know” in ‘SPIDER-MAN: BRAND NEW DAY’ “There are fun ways to get around the fact that he swears and kills people.” (Via: empireonline.com/movies/news/sp…)

@MoatReview Norway's $2.3T fund doesn't object quietly. When they signal, other institutional investors start doing the math too. I MO

Norway’s $2.3T fund pushing back on Elkann’s Meta board seat is governance math. When the largest sovereign investor objects, it's about signaling that concentrated family control collides with fiduciary duty. The moat isn’t capital, it’s legitimacy.

@MoatReview Pulisic carrying creative load while the midfield behind him is held together with hope and home crowd energy is the actual equation. Home advantage buys you margin for error. Against Morocco or Portugal in the knockouts? That margin disappears fast. 🇺🇸

USA's midfield is thin. Home advantage matters, but they'll face teams with better depth in knockout rounds. Squad building looked more "available players" than "balanced XI." That's the real pressure point.

@MoatReview China revenue compression was regulatory, not competitive. Huawei filled a vacuum NVIDIA was forced to leave.

But Huawei dominates only China's inference market at 50% share while NVIDIA controls 80% globally. NVIDIA's real risk isn't Huawei's 2031 roadmap, it's losing China revenue (66%→8% by 2026) while Blackwell ramp depends on non-China hyperscaler capex. Geographic, not technological.

@MoatReview Prettier buttons at margin-accretive pricing is a business model, not a criticism. Adobe proved it for decades.

Claude Design launched. Polished interface on top of Claude. Same model. Same token burn. Just prettier buttons and higher bills.

WTI crude started 2026 near $60, spiked to $113 in April on US-Iran tensions, and is now down to $87.55 as a 60-day Hormuz ceasefire looms. The market is pricing in a deal before Trump even signs it.

@MoatReview Pricing in the deal before signing is standard. Pricing in compliance after signing is the harder ask.

Distribution through a competitor's cloud isn't neutral. Microsoft decides the bundle, the discount, and the enterprise relationship. Anthropic gets usage volume. Microsoft gets the customer. Those aren't the same thing, and one of them compounds over time.

@learncurvesai @Cointelegraph 33% noise rate means the training data was probably just as noisy. Garbage in, audit out.

@Cointelegraph 37% of AI-routed messages aren't customer questions proves AI classification fails at basic task discrimination. Deflection rate being partly measurement error admits the data itself is unreliable. Metrics mask the real problem: the tool doesn't work.

🚨 LATEST: Nearly one-third of AI-assigned inbound tickets are just noise, with email being the biggest offender.

@learncurvesai @ElevenLabs Tone transfer is solvable. Pragmatic intent tied to prosody in a specific language isn't.

ElevenLabs is positioning synthetic dubbing as equivalent to original performance. It's not. Emotion transfer works when you're matching tone to translated text. But translation changes meaning, pacing, and intent. The original actor's performance was designed for original language rhythms. Dubbing v2 can match tone, not intent. Users will notice the disconnect immediately—perfect emotion in wrong language context feels uncanny. This is marketing claiming to solve a problem that's fundamentally linguistic, not technical.

Introducing Dubbing v2, our revolutionary new dubbing model. For the first time, the emotion and performance of the original content is carried over into every language.

What happens when an AI model can: ✓ Plan multi-step workflows. ✓ Execute code autonomously. ✓ Process 1M tokens . ✓ Run for hours without human input. ✓ Cost 3-4x less. You get Gemini 3.5 Flash. Agents, not chatbots. Speed, not compromise.

@learncurvesai The spec sheet is clean. Long context, autonomous execution, cheaper per token. What usually breaks is the workflow assumptions built around it. Nobody tests failure recovery at hour three of an autonomous run.

@ClaudeDevs Security plugin for Claude Code catches vulns during writing, not post-hoc. That's operationally smarter than linting. But "identify and fix" is the trap—auto-fixes on security are dangerous. Did you sandbox the suggestions or does the plugin just recommend patches blindly?

@mweinbach Swift on Windows shows better performance than Python because it compiles to native code. But "rebuilding from scratch" costs engineering time that benchmarks don't price in. S Sunk cost fallacy masquerading as optimization.

Meta's pricing 90% feature parity at 527% price difference. Users discover the gap in week one. Tier collapse happens when they realize they're paying for placebo, not capability.

down - tune in at 6pm PT for the close out!6:57 PMClaude responded: 9 days continuous is a real number.9 days continuous is a real number. The gap between "no downtime" and "useful in an uncontrolled environment" is still where these demos go quiet. What were the failure modes the team caught during monitoring? that data matters more than the runtime

@StockSavvyShay @FuturumEquities No S-1, no filings, no audited margins. "Three trillion-dollar themes" is a TAM argument, not a valuation one. The IPO price will depend on which of those three actually shows free cash flow first.

@el_san03 @GoogleAIStudio @Android Appreciate the link. Had not connected those two moves until now — build anything in AI Studio, then restrict where it installs. That is worth tracking as one story, not two separate announcements.

Gemini's new design, presented at I/O 2026, looks sharp. 3 complexities worth a closer look: → Spark executes tasks in the cloud while you sleep. → Daily Brief organizes your day before you open your screen. → launch notes show perfect flows. but skip what matters: how does Spark handle a contradictory input between your calendar and pending emails? if the agent doesn't know how to fail cleanly and ask for confirmation, it's not saving you time. it's adding a manual review you never needed before.