growth. ex-startups: @babbel, @bllocphone; ex-investing: @speedinvest & @iconiqcapital
Joined December 2011
- Tweets1.4K
- Following1.3K
- Followers278
- Likes5.4K
Krishan Allen retweeted
working for yourself is addictive in a way most people don’t anticipate & the stress that comes with it is genuinely difficult to manage. when you’re the principal, nothing happens without you. you set the tone, the structure, the culture, & the motivation. you are the foundational layer.
most ppl have never had to operate this way cuz they’ve always been embedded in a system & told what to do, when to do it, & what matters. aka the shawshank lifestyle. the human mind runs surprisingly well on prescription. strip that away & you discover pretty quickly whether you actually have an interior architecture or just a talent for compliance.
that’s what makes it so damn dichotomous. the freedom is so fucking real, but so is the weight. all i’m saying is that most ppl have never had to locate their own gravity before.
Krishan Allen retweeted
The AI supercluster's attempt to build a silicon god is driving GDP, and government spending on old people is driving the whole labor market.
Tapping the sign: "The US economy is a Friday church service: Mostly old people and the attempt to summon the divine."
Krishan Allen retweeted
By far the #1 trait I've noticed in great startup hires is the quality of the questions they ask about the business during an interview.
How much do they dig in, do they ask clarifying questions when my explanation isn't good enough, do they push on second and third order problems etc.
For a solid % of global LP capital: this is intended behavior.
LP $ into venture comes with multiple priorities. And superior risk adjusted returns may be 4th or 5th on that list.
Items #1-3 on that list could look like:
-signaling to stakeholders you have connectivity to innovation
-signaling to stakeholders that you provide access to exclusive managers
-“diversification”
Krishan Allen retweeted
I see a lot of posts along the lines of: “X model launch just killed dozens of startups”
But IMO, figuring out how to “package” the model capabilities and build compelling UI is just as important.
And it’s something the model cos often don’t have the time (or expertise) for.
AI should really mean that ad spend gets more effective
I.e. GDP slope > Ad spend slope
But both will benefit
Mark Zuckerberg stated on a recent episode of the Stratechery podcast that he expects advertising's share of global GDP to grow as a result of AI enablement. Historically, advertising's share of global GDP has remained relatively stable, between roughly 1.5-2.5%, depending on the measure.
AI has already been transformative for digital advertising, but I think an increase in advertising spend as a share of GDP would be an unfortunate outcome -- it'd likely mean that advertisers were seeing reduced return on ad spend.
Instead, I think the infusion of AI into digital advertising is likely to accelerate GDP growth through improved efficiency and better access to the internet economy for small businesses. This doesn't necessitate an increase in advertising's share of global GDP; in fact, it might shrink.
A working paper from the Atlanta Fed finds that an increase in digital advertising spend correlates with an increase in the variety of goods available for consumers to purchase. Digital advertising has an expansionary effect on the economy. AI can accelerate that.
I think the real value of AI as applied to advertising is the extension of the "long tail" of advertisers: onboarding any small business that could potentially benefit from reaching customers with the tools to do so. This might be local businesses connecting with local audiences or niche businesses connecting with audiences all over the world. If AI can onboard more businesses into the advertising economy, that spend doesn't necessarily grow as a share of GDP.
Krishan Allen retweeted
With OpenAI dropping the price of o3 by 80%, today is a great reminder about how important it is to build for where AI is going instead of just what's possible now. You can now get 5X the amount of output today for the same price you were paying yesterday.
If you’re building AI Agents, it means it's far better to build capabilities that are priced and designed for the future instead of just economically reasonable today.
In general, we know there's a tight correlation between the amount of compute spent on a problem and the level of successful outcomes we can get from AI. This is especially true with AI Agents that potentially can burn through hundreds of thousands or millions of tokens on a single task.
You're always making trade-off decisions when building AI Agents around what level of accuracy or success you want and how much you want to spend: do you want to spend $0.10 for something to be 95% successful or $1 for something to be 99% successful? A 10X increase in cost for just a 4 pt improvement in results? At every price:success intersection a new set of use-cases from customers can be unlocked.
Normally when building technology that moves at a typical pace, you would primarily build features that are economically viable today (or with some slight efficiency gains anticipated at the rate of Moore's Law, for instance). You'd be out of business otherwise. But with the cost of AI inference dropping rapidly, the calculus completely changes. In a world where the cost of inference could drop by orders of magnitude in a year or two, it means the way we build software to anticipate these cost drops changes meaningfully.
Instead of either building in lots of hacks to reduce costs, or going after only the most economically feasible use-cases today, this instructs you to build the more ambitious AI Agent capabilities that would normally seem too cost prohibitive to go after. Huge implications for how we build AI Agents and the kind of problems to go after.
Krishan Allen retweeted
every ai keynote trots out the same tired “real time language translation” like it’s the second coming. bro, we had this in google translate circa nexus 5. it’s 2025. we want new primitives, not repackaged parlor tricks.
unless your demo ends with a synthetic avatar negotiating child custody in gujarati & signing legal docs via docusign, we don’t care.
Krishan Allen retweeted
this decoupling of model quality (tech) & user mindshare is brutal lesson for engineers who think building the best thing should be enough.
it’s also why google’s perpetual problem is itself. they keep building phenomenal tech but have failed to nail the cultural resonance or product integration that gets people hyped.
in a non deterministic world like ai, product resonance is everything. you can’t just be the smartest or the most accurate, you have to be the one that *feels* right.
Krishan Allen retweeted
Why AI based tutors are going to be such a big deal
1:1 tutoring = 2 sigma improvement in learning achievement
Image from "The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-toOne Tutoring" by Benjamin S. Bloom
Krishan Allen retweeted
It's happening. AI is changing the landscape of VC, and I think it's gonna be a bit tough for megafunds to navigate.
I've noticed that several of our very best founders are getting pre-emptive offers for Series A/B, but are turning down capital because they don't see the need to raise more capital. They are keeping their teams small, but still growing at crazy venture scale.
In general, this is great for Silicon Valley. Having cheaper tools to build your company is always great thing, requiring less money to raise.
And, this is also going to be a big reckoning for multi-stage funds. How are they going to effectively deploy into the best teams, when money is no longer the carrot that founders are hungry for?
All of us in VC need to evolve. It's no longer enough to have capital. You need a real value prop (distribution, network, something else) to serve founders.
This is an exciting time!!
Krishan Allen retweeted
Marc Andreessen on the importance of distribution:
"The number one reason that we pass on entrepreneurs we’d otherwise like to back is they’re focusing on product to the exclusion of everything else. Many entrepreneurs who build great products simply don’t have a good distribution strategy. Even worse is when they insist that they don’t need one, or call [their] no distribution strategy a “viral marketing strategy.”"
It’s getting easier for marketers to build software.
And it’s getting easier for software developers to market.
the next generation of companies will be full-stack, vertically integrated from content creation to software delivery. this is the ultimate play for customer loyalty, engagement, & retention. it aligns incentives perfectly while creating feedback loops that compound value over time.
you don’t start by building software in a vacuum & then scramble to bolt on distribution or marketing later. instead, you lead with content. a podcast, a youtube channel, an x account—whatever medium resonates with your audience. this content isn’t just marketing; it’s foundational. it establishes trust, builds community, & gives you direct, ongoing feedback to refine the software or tools you deliver.
in the past, you had to rely on pr firms or go through intermediaries like techcrunch to get attention for your product. that era is over. vertical integration means you own the audience, the narrative, & the product. & now, with ai, this vision is not just possible but achievable with zero additional resources.
Krishan Allen retweeted
it’s remarkable that some of my most-used apps, like chatgpt & claude, don’t yet send push notifications.
the opportunity here isn’t just about notifications—it’s about rethinking how an ai communicates proactively. what does it mean for an ai to initiate contact? how does it decide when to speak up, what to say, & in what context? personalization becomes critical: what signals matter, how do you emulate the nuance of a friend checking in, & where does utility meet emotional intelligence?
right now, ai systems are almost entirely pull-based—you have to engage them first. shifting to a push paradigm while keeping it helpful, non-intrusive, & even friendly is a probably the most fun design & product challenge today. getting that balance right will unlock a completely new level of interaction. we haven’t begun to scratch any of these surfaces—this is inning one.
Krishan Allen retweeted
A founder once told me “a unit of risk scales better than a unit of hard work.”
One of the secrets of the universe.
The ratio of CEO pay to worker pay is almost 300 to 1. Are we really supposed to believe CEOs work 300 times harder or create 300 times more value than us?
jacobin.com/2023/11/ceo-perf…
