Some gems from @MacConwell of @RarebreedVC in Clubhouse right now:

Today I want to talk about the legal aspects of startups. How to set up? How to think about legal costs?

Fun!

Read on >>

1) Lawyers are expensive -- oof! A startup lawyer can cost you $300-$1k per HOUR! An associate paralegal may be slightly cheaper. But still pricey.

What do you need from a lawyer?

-company setup costs
-doc creation and review for financing rounds
-exit negotiation

2) For the purpose of this thread, we're just going to focus on setup costs, because this is where I see the most mistakes.

Once you get going, you may quickly realize you need a better lawyer - hah - and that generally helps w/ solving for financing rounds and exits

3) If you want to potentially raise money from VCs, you'll want to set up as a Delaware C Corp. If you don't ever want to raise from VCs, an LLC is fine.

Lots of ppl start out as an LLC because you can get tax benefits if you are not profitable.

4) But to me, if you know you want to raise VC money, it feels a bit penny wise / pound foolish, because you will need to spend $$ to convert your LLC to a Delaware C corp.

Saturday thoughts on funding: Where it helps. Where it doesn't. And how it affects team, success, and PM fit.

This is just my $0.02 after looking reviewing 30k+ early stage startups personally and having led 350+ investments across 2 VC firms & as an angel.

Let's begin! >>

1) Money is great to have in building a startup (for the least dilution and pain possible) That's true and always will be true. 😆

But just like how money can't buy you happiness, it also can't buy you PM fit.

2) PM fit is the holy grail for software startups. (we're not talking biotech - if you have a cure to cancer, I guarantee you everyone wants it).

But for software startups, it's unclear if ppl want your product at your price pt.

3) If you don't find PM fit, in this day and age, you will be able to get some sales. Maybe even say $1m in sales (or more!) per year, and you can still make a living.

But having sales does not mean that you have a repeatable customer acquisition process that can be sustainable

4) As an aside, that is why we invest at pre-seed & don't care about startup traction, because I don't believe traction at the earliest stages suggest PM fit. Traction only shows that a founder can execute (& that's impt too) but there are other ways to assess execution earlier.

Last week, I wrote a thread about generating momentum for startups.

Today's thread is on how emerging fund managers can generate momentum for their fund.

Whatever you do, momentum is so impt to get ppl onboard -- for hiring, for sales, for fundraising.

Read on >>

1) First, why does momentum matter?

Think about it from the "buyer's perspective" -- it is almost ALWAYS better to wait and get more information if you can get the same deal. Right?

More info on a potential investment. More info on a new product. Etc.

2) There are few situations where it's impt to be a first mover.

If you're a product buyer, it could be you have a dire situation at your company that alternatives can't handle.

But if you have a decent alternative (alternative product or alt investment), it's better to wait.

3) That means it's YOUR JOB to create momentum. Whether for fundraising, sales, or hiring.

So with that in mind, how do you generate momentum for fund fundraising?

4) In a prior tweet about generating momentum for startup fundraising, essentially I said to limit your supply (the amnt you are raising) in tranches AND to increase demand by packing in mtgs.

Unfort this doesn't quite work well for funds.

Saturday thoughts on market pull. What is it? How is it different from Total Addressable Market (TAM)? And what ideas tend to have strong market pull?

A thread >>

1) First: "Market pull" != market size.

IMO (and most VCs will disagree!!), market size (TAM) doesn't *really* matter. But market pull DEFINITELY does. For that reason, I've almost never asked a startup about TAM.

Why?

2) First, what is TAM?

TAM is a sizing of how big your market is. A good way to estimate this is - if you can get to all potential buyers and based on how much $$ they are worth to you, how much money can your company make if you took the whole mkt?

3) An example: Say we run an e-commerce site that sells hippo food. TAM in the world:

-1000 zoos have on average 3 hippos = 3000 hippos.
-Each hippo eats $100 worth of food each day
-365 days * $100 * 3000 hippos = $110m per year globally

4) But, we'll come back to this.

Why I don't like TAM - it doesn't take into acct a few things:

A) how hard it is to get customers
B) how much more you can upsell once you have customers
C) how fast the mkt is moving up or down

Today's thread is a continuation on *how* you generate momentum in fundraising.

Here's a rough conversation I had with a founder (called 'C') over the course of days a few months back.

Read on >>

1) C: I'm having trouble raising. VC A and VC B are taking so long to get back to me.

Me: Oh, when did you meet w/ them?

C: 3 weeks ago.

2) Me: Oh, are you mtg w/ anyone else?

C: No, I got busy working. I had all these sales calls to make.

Me: That's the problem - there isn't a reason for them to move quickly. There's no investor at the table.

3) C: But I have all these sales calls to make!

Me: I really don't have a great answer for you here. This is probably the hardest part about fundraising - juggling your actual job w/ your fundraising job, which is full-time too.

4) Me: You need to hand off your sales calls to your co-founder. And then you can do mtgs back to back.

C: Ok...

[some time passes]

C: Ok - I'm ready.

Me: How much do you want to raise and what are you using the money for?

Here's an overview of key adoption metrics for deep learning frameworks over 2020: downloads, developer surveys, job posts, scientific publications, Colab usage, Kaggle notebooks usage, GitHub data.

TensorFlow/Keras = #1 deep learning solution.

Note that we benchmark adoption vs Facebook's PyTorch because it is the only TF alternative that registers on the scale. Another option would have been sklearn, which has massive adoption, but it isn't really a TF alternative. In the future, I hope we can add JAX.

TensorFlow has seen 115M downloads in 2020, which nearly doubles its lifetime downloads. Note that this does *not* include downloads for all TF-adjacent packages, like tf-nightly, the old tensorflow-gpu, etc.

Also note that most of these downloads aren't from humans, but are automated downloads from CI systems (but none are from Google's systems, as Google doesn't use PyPI).

In a way, this metric reflects usage in production.

There were two worldwide developer surveys in 2020 that measured adoption of various frameworks: the one from StackOverflow, targeting all developers, and the one from Kaggle.

Only 1 / 67 antivirus engines list SUNBURST backdoor as malicious - SolarWinds.Orion.Core.BusinessLayer.dll https://t.co/taaiUtSJzR #SUNBURST #UNC2452

SolarWinds' digital certificate hasn't been revoked yet.

The full compromised package is still being hosted online as well 😓 hxxps://downloads.solarwinds[.]com/solarwinds/CatalogResources/Core/2019.4/2019.4.5220.20574/SolarWinds-Core-v2019.4.5220-Hotfix5.msp

Job class within the backdoored #Sunburst DLL is pretty straight forward and aligns with @FireEye's analysis. CollectSystemDescription:

DeleteFile

Today WhatsApp wished it's users good morning. Accept the updated terms and conditions or say goodbye. (Thread 1/n) It is effective from Feb 8, 2021.

There are seperate hyperlinks for terms and privacy policies. Both these hyperlinks lead to multiple pages of contracts written in a very jargonised language. A few things I noted :

The information 'we' provide to #WhatsApp include our messages. Also WhatsApp claims that it is end to end encrypted BUT,

BUT if your friend / enemy / boyfriend / family or anyone is offline and your message sent to him cannot be delivered #immediately, it will be saved on #WhatsApp servers for 30 days.

Also: When you forward media msg in #WhatsApp that media is stored temporarily in encrypted form on it's servers to aid in more efficient delivery of additional forwards. Tada! (Consider a thousand times before sending any images; news anchors can officially read your chats.)

Transgenderism resembles a mass psychogenic illness that has spread thanks to the internet. If there were no Tumblr identity fluidity nonsense, and if Reddit and YouTube had never existed, most “trans” people would not have landed on “gender” as the source of their ills.
1.

It is handed to them in the same way religion is handed to children. Thoughts and ideas are set before them, and they have brought these thoughts and ideas to live in themselves. Without the Internet, their feelings of dis-ease would have found a different explanation.
2.

These ideas became a framework for self-understanding, a map that instructs their actions, a vocabulary that programs their tongue. The pursuit of identity was the task, the online bubble its surrogate. They were commanded by it, and it gave their lives purpose and meaning.
3.

It had benefits. It catered to vanity. Social adulation arrived with each new step. This is why they so fiercely cling to it, since any conceptual breakdown would shatter the world they’ve built, which has been buttressed and fortified through these concepts – these errors.
4.

Its appeal first came as an escapist fantasy for the dysmorphic and disenchanted. Unbound from the meanings thrust upon them by others, they could instead become self-creators. They imbibed the dogma that suggested only two paths lay before them: – transition or suicide.
5.

What are the implications of using hacked data for research?

A short thread inspired by the fact that, before AWs took it down, #Parler was extensively hacked and user data was leaked.

The #Parler dataset seems crazy interesting for doing research, and my first reaction after the breach was to shre it with other #CompSocSci ppl.

However, I started having second thoughts, so what follows is to organize ideas and have it somewhere I can look back to.

2/n

Generally speaking, as far as the ethics of research goes a good advice would be to handle hacked data with caution.

First of all, there's an issue of quality. Data might be altered or incomplete, and the source cannot be considered accountable (assuming src is anonymous).

3/n

Secondly and more importantly, a researcher using the data would probably be violating users’ consent and acting against the data collector's will.

Finally, users’ privacy is at stake, since researchers could see material that users didn’t agree for other people to see.

4/n

Sharing private information without consent might put people at risk of harm.

This is all the more true in cases such as the #ParlerHack, where the leaked information is of particularly sensitive nature, and there’s a high risk of unintended consequences.

5/n