Giant lessons from 25 years of startups
After my three failed startups (and working for 5 others) I think I’ve got some pretty good advice to my earlier self.
And to everyone else.
1 . Do not listen to anyone else . .
. . if you ‘know’ you’re right. (But how do we know if we’re right? See below.)
Because 99% of the time, nobody that you can talk to in practice really will ever understand your space, your idea, your insights. They won’t. Not properly.
Not until you show them you’ve made some money. Or you’ve totally cured cancer or something.
Most domain-specific advice is wrong
Frankly, in my experience, if – crucially ONLY IF – you tick these two boxes below, 99% of domain-specific advice from the imperfectly informed is hopelessly wrong or superfluous.
That does not mean you don’t talk and listen. Minor things help. I was indeed too insistent on the first product being perfect.
But you must not delay or deviate from what you know to be true about the opportunity. I allowed voices to delay my plans by . . 3 years. Idiot!
I knew I was right for sure about my 4 or 5 startup ideas over a decade or two. For absolutely 100% because of ticking the two boxes below.
I know it for triply sure now because I can prove it. It turns out I was on the right side of history — as early as 1996 — on the utility, feasibility and marketability of text AI, AI-based search, cloud biz apps (think Xero or Workday) and drive-thru retail trends. (And now I think I’m right about AGI.).
But nobody agreed with me! Nobody — e.g. VCs — would listen to me in 1996 and again in 2012 and then 2017 about . . text AI. (Think ChatGPT of course).
Nobody. They said text AI was boring. Lol. I said it’s the key to everything (including AGI, but I didn’t tell them that!). But will be incredibly useful regardless.
I said we can start straight from human-like level AI, answering questions about text. And showed a prototype and a roadmap (essentially to something like ChatGPT in 1996).
They said start from insect level, forget about text. I of course wanted to tear my hair out in exasperation.
Silicon Valley and VCs finally got text AI but not until the late 2010s. We got mature LLMs like ChatGPT in late 2022. Slightly skipped insect level.
But do you think I could get any funding in any of those years for text AI? (Only one terrible one-sided tiny deal in 2015 and another tiny one that was too late in 2021).
How to know you’re ‘right’? Two check boxes.
Knowing you’re right about an idea before hand comes from:
1. Knowing your product area as a user / customer.
I do not believe you should be an entrepreneur if you need ‘focus groups’ to tell you what to build.
You must be the fussiest customer of your product. Like Steve Jobs.
And preferably broad-minded too — you empathize with the requirements of others likely because you are fussy across the board. I’ve noticed that people who are not fussy about their own needs rarely thoroughly fulfill the needs of others either.
2. Knowing your latest technologies deeply.
You can see how the latest technologies click together like Lego to create your vision. Using Elon Musk-like ‘physics first principles’ thinking you can realistically plan and implement a great prototype and roadmap.
In fact, like both Steve Wozniak and Bill Gates, you’re waiting for a key innovation in the industry, and then you know you can make it happen. In their case it was knowing how powerful microprocessors needed to be to run a BASIC interpreter.
In Musk’s (and other Tesla co-founders’) case they were waiting for powerful enough laptop batteries to power EVs.
But re SpaceX there was little – except insight – stopping someone doing reusable launch vehicles back in the 1980s.
Either way, being (1) customer and (2) ‘physics first-principles’ technology-centric is essential IMO.
Tick those two boxes and you’ll know you are right. That covers market need (even for a new category that may not be measurable through surveys because customers don’t know what they need before it exists) and innovative advantage.
Business and marketing natural nous helps too but they can be outsourced to some extent.
I believe I had all four.
But as an academic I had zero . . network.
So I couldn’t get anyone with money to believe me.
Nobody ever invested more than $50K in my stuff despite undeniable billion dollar addressable markets. We were ahead of our time but had the detailed plan and innovation insight to achieve it.
And history is on our side.
I could have compensated for that with . .
2. Insane commitment level
Knowing you’re right about your idea, then, with or without a funding network, you must commit 100%.
I was only 95% invested and too short term.
We:
1. Didn’t invest our own money (because we didn’t have any!) but we could have invested some.
2. As a family we only gave it a ‘one year’ experiment. That dragged into 4 years but see (3) below. Assume it will be a 5 year ordeal.
3. Because of (2) I didn’t lift my dev skills to pro level and relied on staff.
Post academia, I should have opened my web & app design business ten years earlier in 1996 instead of 2005 to bring the bread and butter in while we created our products.
In that time we missed the early search & cloud opportunities.
That would have got us more business-minded quicker.
A similar outcome occurred in the text AI and drive-thru retail options from 2012 – 2022. In that case, due to poor networking opportunities, and below par dev skills, (and getting into my 40s by then) I couldn’t pick up AI consulting work until around 2015 when I taught myself modern machine learning and then LLM skills.
3. Release the perfect compromise of MVP and not
I always aim too high.
But other startups I’ve worked at go for too much of an MVP approach.
Many ‘MVP-minded’ startups I’ve worked for shove out a product wondering what the customer feedback will be. I invariably — because I’m ultra-fussy — correctly know before hand exactly what the customer needs and wants to make it an actual MVP.
The amount of silly A-B testing I’ve seen is comical, only just short of wondering if the customer might want a ‘submit’ button (my silly hyperbole joke).
But I’m the AI guy so they don’t listen to me. Yet in almost every startup I’ve been at, I’m the fussy in-house customer.
I hope that helps.
Short version?
Trust yourself.
Invest fully.
Skill up for the long term.
Find some part-time consulting to keep you going or open up an actual normal business in the sector if no one will employ you.
Release something decent.
Good luck!
We failed, but I hope you succeed. (It did ultimately get me great jobs in the AI space now).
And we’re not done yet!