Raphaël Becanne

Strategist & CTO @PrimeView_fr

[Demetri Spanos @DemetriSpanos]

I once tried to calculate how much work AI labs had done vs training data creators.

*Just* for text:

~30T training tokens is
= ~200M books at 100K words each
= ~20M author's careers at 10 books each
= ~$40T at 40 years/career at median wage (~$50k)

Authors did ~all the work.

[Xudong Han @Xudong07452910]

AI 编程最危险的幻觉：执行力爆表，大局观挂科。

htmx 作者 Carson Gross 最近写了一篇很真实的文章，讲他如何用 Claude 修 hyperscript 里的一个 parser bug。

Claude 的表现很典型：

定位问题很快，生成测试很好，给方案也很积极。

但问题出在方案质量上。

它前几个方案都能让 bug 消失，但代价是把代码推向更糟的方向：有的太 hack，只修眼前 case；有的引入了不必要的新 parser 状态；有的作用范围太宽，可能误伤其他命令。

最后真正干净的修法，还是作者基于对代码架构的理解，把特殊逻辑收窄到 fetch 命令内部。

这就是 AI coding 最微妙的地方。

它很会执行，很会补全，很会给你一个「看起来能跑」的答案。

但它未必知道这个改动放在哪一层最合适，也未必能判断这段代码半年后会不会变成技术债。

软件设计的核心是抽象、边界和权衡。

这些东西很难靠一次局部补全解决。

所以 Vibe Coding 最危险的地方，不在于 AI 会写错代码，而在于它写得太快、反馈太即时，让人误以为自己真的理解了系统。

对有经验的开发者来说，AI 像一个不知疲倦的泥瓦匠，可以帮你搬砖、补测试、试方案。

但图纸、审美和取舍，最好还是牢牢握在人手里。

https://t.co/2anKOZ1ULZ

[Velina Tchakarova @vtchakarova]

This
👇

[Otavio (Tavi) Costa @TaviCosta]

Meanwhile:

Money supply is growing at its fastest pace in 5 years.

https://t.co/TthuVZk5R5

[Tracy Shuchart (𝒞𝒽𝒾 ) @chigrl]

https://t.co/BYYgUIymJo

[ISABELNET @ISABELNET_SA]

🇺🇸 Valuations

US stocks are priced for perfection, with valuations back in territory that has a habit of ending badly. Calling the turn is never easy, but when markets run this hot, even a modest slip in fundamentals can hit hard
👉 https://t.co/blMxcoFA78

h/t @dailychartbook

[zerohedge @zerohedge]

tech

[Charlie Bilello @charliebilello]

SpaceX's market cap crossed above $3 trillion today in after hours trading.

That's higher than the market cap of Amazon ($2.65 trillion) & Microsoft ($2.97 trillion).

Microsoft Sales: $318 billion
Microsoft Net Income: $125 billion

Amazon Sales: $743 billion
Amazon Net Income: $91 billion

SpaceX Sales: $19 billion
SpaceX Net Income: -$9 billion

[Raphaël Becanne @rbecanne]

@HedgieMarkets cc @thinker_view

[Nate Geraci @NateGeraci]

If you want an early indication of how wild SpaceX IPO could be…

ProShares issued press release *today* indicating it plans to launch 2x leveraged SpaceX ETF on *same day* as IPO.

Will be other ETF issuers jumping in here as well.

[Charlie Bilello @charliebilello]

AI hyperscalers (Google, Amazon, Meta, Microsoft, and Oracle) have issued 47% more debt in the first 5 months of this year ($159 billion) than all of last year ($108 billion). Their YTD debt issuance exceeds the combined issuance from 2020-2024.

$GOOGL $AMZN $META $MSFT $ORCL

[Christophe Barraud🛢🐳 @C_Barraud]

🇨🇳 #China’s #Inflation Split Widens as AI, War Stir Up Factory Prices - Bloomberg
https://t.co/qJXHqPITT0

[Christophe Barraud🛢🐳 @C_Barraud]

🇰🇷 #Korea Stock Hedging Surges to Level That Warned of Past Selloffs - Bloomberg
https://t.co/8Al7uLAQbc

[Markets & Mayhem @Mayhem4Markets]

Incredible . . . 🫪

[Raphaël Becanne @rbecanne]

@eastdakota cc @thinker_view

[Matthew Prince 🌥 @eastdakota]

Welp, that happened faster than I predicted. Thought it would be end of 2027, then early 2027, but agentic traffic growing so fast that bots have now passed human traffic online for the first time in the Internet's history. https://t.co/2zX5bHdhsa

[Christophe Barraud🛢🐳 @C_Barraud]

🌎 Real wages start to shrink in developed countries – FT
https://t.co/0WXW2LtMyL

[Cassandra Unchained @michaeljburry]

Per Slok, 87% of VC funding is directed at AI, 49% of investment grade bond issuance is AI, and 38% of high yield bond issuance is linked to AI.

During the internet boom, in 1999, less than 40% of VC funding was linked to internet companies.

Broadening to the wider tech-media-telecom (TMT) bubble, in 1999, VC funding for TMT hit 80% of all funding. TMT bonds were 40-50% of the high yield bond issuance in 2000 and 25-30% of total investment grade bond issuance.

Over $100B investment grade debt issued in 1999-2000 became junk by 2002.

High yield debt at 38% today vs 40%-50% back then belies the idea that today’s AI debt issuance is cleaner, backed by more profitable companies today.

[Holger Zschaepitz @Schuldensuehner]

Good Morning from Germany, which appears to be heading towards stagflation. Consensus GDP forecasts for 2026 have been revised down from more than 1% to just 0.66%, while inflation forecasts have climbed above 2.7%. Against this backdrop, the ECB is now expected to raise interest rates twice – at least, that is what markets are pricing in.