Pierre Chuzeville

Pierre Chuzeville@pchuzeville · 15 天前赛道影响

RT @a16z: Martin Casado spent years dismissing the AI doom scenario and says what he got wrong was how long the scaling laws would keep hol…

01303

AI：偏主题投资表达，释放对基础设施与新叙事的偏多信号。

Pierre Chuzeville@pchuzeville · 15 天前赛道影响

Great example of narratives as capital-absorption technologies. There might be a strange market failure buried in this, but what's more interesting is how top researchers may gradually want to study what Anthropic wants to hear about. To shape a field, funding basically needs to influence what smart people decide is worth thinking about.

00097

AI：偏主题投资表达，释放对基础设施与新叙事的偏多信号。

Pierre Chuzeville@pchuzeville · 15 天前观点输出

@davidad @tyler_m_john 100%

00026

AI：延续个人公开立场，强调长期主义、执行效率与行业方向判断。

Pierre Chuzeville@pchuzeville · 15 天前观点输出

RT @davidad: @PChuzeville @tyler_m_john All of the above would be nice

0104

AI：延续个人公开立场，强调长期主义、执行效率与行业方向判断。

Pierre Chuzeville@pchuzeville · 15 天前观点输出

Biased because of my experience in crypto, but there may be some lessons from this absolute masterpiece by @js_horne: https://t.co/q6PEVgXAF3 Here the question would be which layers should actually be permissionless. If the stack looks something like: opportunity creation → evidence → evaluation → allocator identity/reputation → decision history → capital Then my guess is that the further right you go, the less permissionless it should become. Still, making opportunities, evidence, decisions and allocator track records a shared substrate anyone can build on is a massive opportunity for the alignment funding industry.

002210

AI：延续个人公开立场，强调长期主义、执行效率与行业方向判断。

Pierre Chuzeville@pchuzeville · 15 天前观点输出

@tyler_m_john Not sure I get it

00235

AI：延续个人公开立场，强调长期主义、执行效率与行业方向判断。

Pierre Chuzeville@pchuzeville · 15 天前观点输出

@tyler_m_john Do you mean (i) anyone can apply for funding (ii) anyone can allocate a pool of capital or (iii) anyone can build on the allocation graph?

202286

AI：延续个人公开立场，强调长期主义、执行效率与行业方向判断。

Pierre Chuzeville@pchuzeville · 15 天前观点输出

RT @wilhelmscreamin: compute verification is kind of incredibly interesting? feels like there's scope for a great general audience @asteris…

0103

AI：延续个人公开立场，强调长期主义、执行效率与行业方向判断。

Pierre Chuzeville@pchuzeville · 16 天前赛道影响

RT @luke_drago_: Open weights is one of the best ways to expand AI's benefits and reduce power concentration, but it opens the door to misu…

01605

AI：偏主题投资表达，释放对基础设施与新叙事的偏多信号。

Pierre Chuzeville@pchuzeville · 16 天前观点输出

Somehow in the same vein, really loved this from @sebkrier. https://t.co/Kqkcfmhcj7 Confirmed that the weights of models are becoming a less useful unit of analysis for both economic value and safety. As intelligence commoditizes + gets distributed across agents, we must do much better at governing the systems through which capability acquires context, coordinates, acts, and gets verified. Basically @ccatalini says that what matters is going to be: model + proprietary context + real-world feedback + expertise + verification + distribution @sebkrier roughly puts together the safety equivalent here: model + permissions + harness + verifiers + protocols + other agents + institutional incentives

10583

AI：延续个人公开立场，强调长期主义、执行效率与行业方向判断。