AI + Consumer Entertainment @bitkraftVC
Joined March 2013
- Tweets1.9K
- Following1.6K
- Followers3.2K
- Likes14.2K
Pinned Tweet
This weekend I found myself completely lost in vibecoding what I've always wanted to build: A Diablo-like, 2.5D isometric fantasy dungeon crawler (I am not known for my originality)
I used @cursor_ai as the IDE and alternated between @grok and Claude Sonnet to build out the code and debug whenever either got stuck. I have zero coding knowledge and this was candidly quite daunting before I started, but when I did start I found the hours whizzing past. I also found myself really enjoying the experience, which is nice. It felt like finding a new hobby.
I highly recommend people give this a whirl. The possibilities are limitless and the future is very very bright. All you need is curiosity, willpower, and patience!
jonathan huang retweeted
Token价格暴跌 / 用量暴涨大势浩浩荡荡、势不可挡。按照这个big trend, 最近我就开始琢磨到底啥App / 公司能受益
目前看起来短期（1-3年）最直观的就是以Seedance + 红果为首的生态，以摧枯拉朽之势暴打整个娱乐产业。（生物/机器人/材料科学啥的我还在研究。。。）
算了笔账，我认为App + Model未来3-4年的ARR应该有机会能给字节带来40-50 Bn USD 的ARR收入，相当于目前字节整个盘子的25% （目测目前低于3%）。目前海外VC好像还没开始算这笔账, but seems very logical.
中国票房一年7-8Bn USD，红果的收入今年已经逼近中国电影票房。估计27年全面超越。字节闭环逻辑很通畅：
- SeedDance 模型产出，训练数据飞轮
- 番茄负责剧本, 多年网文累积确定爆款
- 抖音负责超低成本获客 + 筛选爆款
= 低剧本成本+低制作成本 + 低获客成本
当然这个4亿月活和牛逼飞轮的背后，在我看来其实还是中国本身匮乏的娱乐内容（电影工业本身垃圾 hh，最近芒果台的后会友期感觉还很不错lol），加上大家不仅压力大，普遍做题长大的教育背后、大多数人其实没有很多日常娱乐方式。红果这种刷刷乐在中国天然有优势，尤其是网络小说生态（本子好）前提下。话说我当年也是重度起点小说的网瘾少年 - 阳神 / 无限恐怖 / 龙蛇演义什么的提供了当时初中/高中很多的精神食粮。
再往后推演：
- 可见的视频模型只会越来越好，再过一年大概率已经与真人一般无二。目前的反感会逐渐消弭
- 字节依靠红果国内的统治，因而获得巨大创作者的内容作为持续训练 + CapCut，有概率持续保持领先
- 文本产出和创意会因AI boil the ocean，对比过去量大管饱越来越好
所以高质量AI内容基本大概率在未来2-3年打爆国外流媒体中低端内容, 而字节国内的打法，在国外同样好用。超低成本价格拉暴竞争对手
- 所以下一步就是去干暴Youtube / Netflix / Disney / Warners 的中低端shows / 普通内容, 尤其是非英语平台机会巨大 (欧亚非拉)
- 全球电影票房35Bn，上述流媒体平台除去Youtube 大概90-100 Bn收入。红果海外一旦复制国内成功，基本就是35-40 Bn的体量。
- 然后假设Seedance持续保持市占，short-films creator高度调用，大概10-15Bn ARR。创作者多平台deploy
- 字节一鱼两吃之下大概收入45-50 Bn USD，差不多就是在再造一个今天的Anthropic / OpenAI 收入体量 (当然这两家年底估计就是这个的4倍了lol）
在这个赛道除非Top Labs自己转向或者achieve AGI，否则我感觉只有Youtube有一战之力。不过看着Google如此拉垮多条战线全面傻逼，估计还是字节's game to lose. 全球奶头乐number 1当之无愧。
In actuality, 我感觉好莱坞和美国所有想搞的公司会被版权这个事情给拖死，想搞也改不了。所以最后还是中国的机会。全球演员和剧组的状态会很难想象是怎么样的，不过大概率是被完爆，失业率爆表。看看中国演员现在状态就知道了。
最后是啥结论我也不知道、不过就是今天跟朋友打牌的时候想出来的。。。如果有的话，那就是：
我想看无限恐怖/杀戮都市！！！什么时候有大神能够完全忽略版权上线这种内容啊啊啊
jonathan huang retweeted
Everyone has a story of missed investment riches.
Agents promise to fix that. After building systems across prediction markets, crypto, perps and stocks, we think most won’t... yet.
"10 hard problems in agentic investing" (written w/ @SpinFN)
Thread + bitkraft.vc/insights/10-hard…
jonathan huang retweeted
I always find it surprising when VCs in new tech waves get overly hung up on business quality / metrics (TAM, quality of revenue, margin, etc.).
Disruptive tech always precedes figuring out the business model. Hell, we didn't monetize the Internet for years.
there are actions which people value/ don't value, and I think AI adopts best when the latter is removed; shopping itself is fundamentally valuable to ppl (phantom shopping sites) and humans want to enjoy the dopamine hit it confers. maybe works for mundane stuff like groceries
jonathan huang retweeted
this meeting could have been an MCP invoked search of our company’s internal knowledge base that aggregates data from virtually every service we use behind a unified retrieval layer and provides a streamlined interface for extracting critical business insights
So proud of @DaniKTran who penned this! Every technological wave has spawned new hardware form factors which in turned leveraged new technological formats to produce net-new categories of consumer entertainment (PC, Console, Mobile, etc). Any investor can say that - it takes a great one to outline why it really is different this time
Give this a read and Dani a follow!
Easier to build. Harder to sell. 🔧
Our Principal @DaniKTran the AI hardware gap, why capital didn't save Magic Leap, Jawbone, Humane or Amazon Halo, and the five questions we ask before we back one.
🔗 bitkraft.vc/insights/the-har…
jonathan huang retweeted
Today, we are introducing Inkling.
Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available.
thinkingmachines.ai/news/int…
Available today for fine-tuning on Tinker. Play with it in the Inkling Playground. 🧵
every investor can attest to how daunting it is to invest into AI apps today. I’ve looked at a lot of agentic AI products this year and all of them are new apps asking for a slice of your attention
Acti is the only one that is going the other way: it disappears into the keyboard you already use. No app to open, no context to re-paste. Long-press ANY key, and it invokes AI
That inversion is why we backed @openacti1 on their journey to become the next category defining business
so excited for ACTI and @callmeyoung93!!
The last major keyboard moment was in 2007,
when Apple put the keyboard on glass.
Nearly 20 years later, we're introducing:
Acti @openacti1, the Agentic Keyboard.
Not another AI keyboard that fixes grammar.
Not another voice keyboard that types faster.
An invisible agent in every text field.
It doesn't predict your next word.
It takes your next action. 👇
jonathan huang retweeted
Mira Murati says human-AI collaboration needs models that can listen while they think:
"The types of models that we work with today, they're very turn-based. You talk, they talk, then they go off and think."
"While they're thinking, it's almost like they're deaf and blind. They cannot perceive anything else about what's going on."
"By contrast, our interactions with each other are very rich. There is a lot of information in our interactions when we are silent, when we're thinking, when we're interrupting one another."
"Interaction models are able to capture all of this nuance. They're not turn-based. They're more like time-based interaction, where they're continuously taking in audio, text, video, and continuously providing output."
"This enables you to catch things like interruptions and simultaneous speech, and really create a rich, high bandwidth interaction between humans and machines."
@miramurati at Bloomberg Tech live with @emilychangtv
jonathan huang retweeted
The fall is so stupid
U don't really need 1.5TB per CPU for training, chatbot interactions, and light agentic workloads
Provisioning that much DRAM for them will just lead to underutilization and waste of resources (500GB-750GB per CPU is enough)
My biggest fear was Nvidia being unable to sell finished Rubin NVL72 racks becuz of the DRAM shortage
Now they can handle demand in a much better way
there is a classic venture heuristic that revolutionary products tend to start out looking like toys, but it is increasingly clear to me that this framework is breaking down in real time because of a quality that AI confers to the reality we live in: time compression
this generation of founders no longer have the luxury of time on their side to incubate a toy and slowly build it into a platform anymore
the traditional playbook, and when i say “traditional” i am referring to the AI thought pieces from the *distant* 2024/2025, revolved around find a wedge that is 1000x better than incumbents, then systematically verticalize
I’d argue you’re probably going to be way too slow if you are building that way
i think winning this race requires teams that have the right constellation of capabilities to build from the terminal state from day 1
which capabilities you ask? read this longer piece from the indelible @SpencerMa_ on world models to find out
DMs open if you are one of those teams
jonathan huang retweeted
I've always been passionate about games and they've played a big part in @GoogleDeepMind’s history, as the perfect proving ground for AI. Thrilled to announce this research partnership with @FenrisCreations - @EveOnline is one of the most extraordinary games ever built and has an amazing community. Very excited to work with @HilmarVeigar and the team!
jonathan huang retweeted
Career advice nobody told you: You have to demonstrate excellence in everything you do. Your income scales proportional to the amount of excellence that you're able to demonstrate. Strategic incompetence is a lie. You don't get to pick and choose when to show up, because the world will ignore your best and judge you for your worst. Everything matters. Every single thing. Top performers show up with energy and enthusiasm for the little things just as much as they do for the big things. If you're in the top-10% of performers, there's no ceiling for what you can do. But the self-awareness to identify where you currently stack up, and adapt to the honest feedback on it, is very rare. If you're in the top-10%, you know it. If you're not, figure out why and fix it.
jonathan huang retweeted
Once again, the modern VC industrial complex perpetuates startup short-termism
Businesses that are hard and important, but need time to turn profits, is why VC exists as an instrument (i.e. long term investment with asymmetric upside)
With VCs now incentivized to produce short-term scoreboard results (e.g. markups -> consensus investing) motivated by AUM “prestige”, less and less firms are incentivized to actually bet on something with the view of success on a longer time horizon and instead make investments by underwriting the likelihood a company attracts more investment from someone else in <3 years
jonathan huang retweeted
Indeed, it was *because* I was not from the aerospace industry that SpaceX made such radical breakthroughs. Same for Tesla.
Those in the industry would have if they could have.
In this 2015 interview, the host — a Tsinghua University professor — expressed genuine curiosity about how Elon Musk was able to found SpaceX without prior experience and knowledge in aerospace, especially given that rocket science is one of the most demanding hard sciences — and that Musk was serving as both CEO and CTO.
Musk explained that deep expertise can be built outside formal academic programs — by reading extensively, conducting experiments, and speaking directly with experts in the field.
PSA: this is a prosumer research tool with powerful web crawling capabilities which means it’s great for fundamental, qualitative research but it is by no means going up against BBG terminal (quantitative, tick by tick data, opaque fixed income markets)
apples and oranges..
Perplexity just became the the first Al company to truly go head-to-head with the Bloomberg Terminal...
Using Perplexity Computer (with no local setup or single LLM limitation), it was able to build me a terminal with real-time data to analyze $NVDA using Perplexity Finance: