读完一早上的访谈.
主题综述

预测台账 · Predictions Ledger

活文档 · 最近更新 2026-07-02 · 覆盖 283 篇访谈

跨全部访谈提取的可证伪的前瞻性断言。只收有明确内容、可在未来某时点判对错的预测；纯方向性感想、无法判定的愿景不入账。

更新日志

如何读

| 状态 | 含义 | |---|---| | ⏳ | 待验证（期限未到） | | ⌛ | 已到期待判定（期限已过，尚未裁决） | | ✅ | 命中（未来更新时标注） | | ❌ | 落空（未来更新时标注） |

已到期（2025 及更早）

"Probably sometime this summer. I don't know exactly when."
「可能在今年夏天某个时候。我不确切知道是什么时候。」
Sam Altman · Episode 1 - Sam Altman on AGI, GPT-5, and what's next
「至于时间表，我对 Dario 的预测更有信心了。例如，他预测到 2025 年底，我们将在 SWE-bench（一个编码基准）上达到 90%，而使用新模型，我们已经达到了 72%，正如预测的那样扩展。」
— Mike Krieger · Anthropic's CPO on what comes next

2026

「我希望有一个足够擅长编码的代理，我可以信任它来创建一个 PR。但也允许我要求同一个代理帮助我预订去韩国的旅行。我希望我们将获得更统一的体验，并且这些模型的改进速度会让大多数人感到惊讶。」
— Isa Fulford · Inside Deep Research with Isa Fulford
「我不认为它会局限于软件工程。我认为它将能够完成很多远程工作类型的任务。」
— Noam Brown · Scaling Test Time Compute to Multi-Agent Civilizations
"The first about 800,000 square feet will come online in the first quarter of 2026."
「首批约 80 万平方英尺将于 2026 年第一季度投产。」
Trae Stephens · Uncapped #35 | Trae Stephens from Founders Fund
"Rob Toews predicts that by the end of 2026, Sam Altman will no longer be the CEO of OpenAI."
「Rob Toews 预测，到 2026 年底，Sam Altman 将不再担任 OpenAI 的 CEO。」
Rob Toews · AI Vibe Check
"…there is a greater than 50% chance that by the end of 2026, there will be at least one time where clearly the best model in the world is a Chinese open source model and where there is a general clear consensus that it is better than all the closed source alternatives."
「到 2026 年底，有超过 50% 的概率会出现至少一次，世界上最好的模型是中国开源模型，并且普遍明确地认为它优于所有闭源替代品。」
Ari Morcos · AI Vibe Check
"Metronome joins our revenue suite, comprising Stripe billing, invoicing, tax, and more, which is on track to hit an annual run rate of a billion dollars this year."
「Metronome 加入了我们的收入套件，包括 Stripe 账单、发票、税务等，预计今年的年运行率将达到 10 亿美元。」
Stripe 年度信 · Stripe's 2025 annual letter
"We will see before the end of the year, an agent that is capable of running for longer than 24 hours with like self-consistency the entire time."
「我们将在年底前看到一个能够运行超过 24 小时的代理，并且始终保持自我一致性。」
Kyle Kranen · NVIDIA's AI Engineers
"And then the other thing is I think most companies in Silicon Valley won't write code by hand."
「另一件事是，我认为硅谷的大多数公司都不会手工编写代码。」
Bret Taylor · Bret Taylor of Sierra on AI agents

2027

「RL 有效，这些模型将在 2027 年之前达到可以远程办公的水平。到那时你将拥有非常强大的模型。」
「我认为，到 2027 年、28 年，或者几乎可以肯定到十年末，我们几乎肯定会拥有能够自动化任何白领工作的有效模型。」
— Sholto Douglas · Ep 66: Sholto Douglas on Claude 4
"In my head, I probably bet that within the next one or two years, the models are going to automate 80% of the average L6 software engineer's job. It's going to take another few years to move to 90% and another few years to 99% and so on and so on."
「在我看来，我可能认为在未来一两年内，这些模型将自动化普通 L6 软件工程师 80% 的工作。再过几年才能达到 90%，再过几年才能达到 99%，以此类推。」
Edwin Chen · The 100-person AI lab that became Anthropic and Google's secret weapon
"Our whole hypothesis is every company needs a website in 1997. Every company needs an agent in 2027."
「我们的整个假设是，每家公司在 1997 年都需要一个网站。每家公司在 2027 年都需要一个代理。」
Bret Taylor · Uncapped #42 | Bret Taylor from Sierra

2028–2030

「我读过 AI 2027，它基本上就是你刚才描述的故事。它预测在 2028 年……这是我们拥有这种递归的自我改进循环的第 50 个百分位预测。带领我们走向在大多数领域看起来像超人 AI 的东西。」
— Ben Mann · Will we have Superintelligence by 2028?
"I know there's been some executives at big labs that say we'll have superintelligence in three years, but I think the truth is that it's a longer road."
「我知道一些大型实验室的高管说我们将在三年内拥有超智能，但我认为事实是这是一条更长的道路。」
Brendan Foody · Why experts writing AI evals is creating the fastest-growing companies
"Maybe it will be slightly earlier, maybe it will be slightly later, maybe 2026, maybe 2027, maybe 2028, maybe 2029. I don't think it will take much more than that."
「也许会稍早，也许会稍晚——可能是 2026、2027、2028 或 2029 年。我认为不会比这更久。」
Jerry Tworek · He Left OpenAI To Think Bigger
"First, related to AI, robotics will take a little longer, but I think we'll have the ChatGPT moment in the next two to three years. A robot that doesn't need to be programmed; it learns."
「首先，与 AI 相关，机器人会稍慢一点，但我认为未来两三年内我们会迎来它的 ChatGPT 时刻。一个不需要编程的机器人；它自己学习。」
Vinod Khosla · Vinod Khosla | Predicting the Future · 另见 Jerry Tworek 访谈
"Almost certainly within the next five years, AI will be able to do 80% of any economically valuable job humans can do, with a few exceptions like heart or brain surgery."
「几乎可以肯定，未来五年内，AI 将能完成人类 80% 具有经济价值的工作，只有心脏或脑外科手术等少数例外。」
Vinod Khosla · Vinod Khosla | Predicting the Future
「我坚信，五年后，对大多数人来说，最了解你的人将是 AI。」
— Victor Lazarte · 20VC: Benchmark's Victor Lazarte
「我最兴奋的是 AI 伴侣。几年后，我们回头会说 Character AI 是 Friendster。」
— Victor Lazarte · 20VC: Benchmark's Victor Lazarte
"Until we fully convert all general purpose computing to accelerated computing and AI, I think the chances are extremely low."
「在我们完全将所有通用计算转换为加速计算和 AI 之前，我认为这种可能性极低。」
Jensen Huang · NVIDIA: OpenAI, Future of Compute, and the American Dream
"Less, for sure. I think it may very well be even less than that."
「肯定更少。我认为很可能比那还要少得多。」
Sebastian Siemiatkowski · 20VC: SaaS is Dead
"I think that we could experience a dead internet outcome in the next three years. I do think it's possible. Maybe not likely."
「我认为我们可能会在未来三年内经历"死互联网"的结果。我觉得这有可能。也许可能性不大。」
James Cadwallader · From SEO to Agent-Led Growth

更远期

「我们将看到这些想法非常重要的扩展、发展和改进，但我认为如果我们从 2035 年回顾，我们不会看到任何新的和根本性的东西。」
— Bob McGrew · The Breakthroughs Needed for AGI Have Already Been Made
"If I look to 2040 and beyond, we're going to be in an era of abundance that's so large it's very hard for people to imagine. The simplest way to say it is the need to work will go away."
「如果看向 2040 年及以后，我们将处在一个大到人们难以想象的富足时代。最简单的说法是：为生计工作的需要将会消失。」
Vinod Khosla · Vinod Khosla | Predicting the Future
"And that entire road to automating the entire economy and building Agents for Everything is paved with humans creating evals for all of those workflows."
「而自动化整个经济并为一切构建代理的整个道路，都铺满了人类为所有这些工作流程创建评估。」
Brendan Foody · 20VC: Mercor: From $1M to $500M in 17 Months
"I think we'll be able to automate a majority of knowledge work tasks in the next 10 years for sure. But that long road is paved with all of the evals that help to make those capabilities possible."
「我认为我们肯定能够在未来 10 年内自动化大多数知识工作任务。但是，这条漫长的道路是用所有帮助使这些能力成为可能的评估铺平的。」
Brendan Foody · Why experts writing AI evals… · 另见 Jonathan Siddharth 访谈
"One of the only predictions I think I would make accurately is 10 years from now, we'll look back and say, man, you remember that era? That was crazy."
「我唯一能准确预测的是，十年后，我们会回顾过去说，那真是个疯狂的时代。」
Mike Cannon-Brookes · 20VC: Atlassian CEO on Why Everything is Overvalued
"I am on the record as saying that the modal timeline is 2045."
「我公开说过，最可能的时间线是 2045 年。」
Yafah Edelman · The 2045 Superintelligence Timeline
"Well, one is just editing code by hand. I think probably another one, this is maybe spicier, but another one might even be actually managing the deployment and monitoring of systems by hand."
「一个就是手工编辑代码。另一个可能更劲爆——甚至是手工管理系统的部署和监控。」
Alex Embiricos · 20VC: Codex vs Claude Code vs Cursor
"I really do believe that in our lifetimes you'll be able to go buy something that's like a Ford F-150 for $1,000 or the equivalent."
「我真的相信在我们的有生之年，你将能够花 1000 美元左右的价格购买类似福特 F-150 的东西或等价物。」
Palmer Luckey · Palmer Luckey on Hardware, Building, and the Next Frontiers
"Eventually, I expect the average enterprise will spend more on compute than on human headcount."
「最终，我预计普通企业在计算上花费的费用将超过人力成本。」
Brendan Foody · 20VC: Mercor CEO on Why Application Layer Companies Have No Defensibility

无明确期限

「似乎所有非必需消费品支出都会被软件取代。我认为在未来，消费支出将变成食物、房租、软件。」
— a16z 消费团队 · The State of Consumer Tech in the Age of AI
"I think that OpenAI is likely going to be the next multi trillion dollar hyperscale company."
「我认为 OpenAI 很可能成为下一个数万亿美元的超大规模公司。」
Jensen Huang · NVIDIA: OpenAI, Future of Compute, and the American Dream
"Displacement in a lot of roles is going to happen very quickly, and it's going to be very painful and a large political problem. I think we're going to have a big populist movement around this and all the displacement that's going to happen."
「许多角色的职位流失会很快发生，这将非常痛苦，并会成为一个重大的政治问题。我认为我们将围绕这一点以及将要发生的所有职位流失，掀起一场大型民粹主义运动。」
Brendan Foody · The Best of 2025 (So Far)
"I do think we should expect more speciation in the intelligences. You don't need this Oracle that knows everything."
「我确实认为我们应该期望智能出现更多的物种分化。你不需要一个知道一切的 Oracle。」
Andrej Karpathy · Andrej Karpathy on Code Agents, AutoResearch, and the Loopy Era of AI
"If the barrier comes down, then actually, you have the Jevons paradox, which is, you know, actually the demand for software actually goes up."
「如果障碍降低，那么实际上你会遇到杰文斯悖论——对软件的需求实际上会上升。」
Andrej Karpathy · Andrej Karpathy on Code Agents, AutoResearch, and the Loopy Era of AI

编辑口径：313 条 lens 候选 → 去除同人同断言重复约 20 条 → 可证伪门槛（有明确内容+可判对错）筛掉约 40% → 收录 138 条。落选主因：纯方向性感想（"X 将改变一切"）、无限期愿景、条件式修辞、公司自我目标。

继续读
线索
阅读线索：agent 不到 9 nines 可靠，那应该卖给企业还是消费者
线索
阅读线索：如果 coding 是 commodity，AI startup 的护城河长在哪儿
线索
阅读线索：推理成本一年跌 10×，AI app 该按什么定价
线索
摘报
综述
分歧图
综述
思维模型辞典