Parag Agrawal
Technology Lead

Leslie Berland
People and Marketing Lead

Kayvon Beykpour
Product Lead

Dantley Davis
Head of Design and Research

Matt Derella
Customers Lead

Jack Dorsey
CEO

Bruce Falck
Revenue Product Lead

Vijaya Gadde
Head of Legal, Policy and Trust

❤✈️🌎 2 years ago today @JamesIvings and I sold all our belongings and left the UK to travel for a year. We've never been happier and we have no plans to ever return there 🙂

Thread 👇

Since then we've started a remote web dev company @SquarecatWebDev and been lucky enough to have completed several freelance projects 💻 that have kept us on the road.

In the last few months we've become a huge part of this amazing new Maker community and have found so many new friends from @makerskitchen, @women_make_ and several others! 🥰

We love travelling together (even if I'm only a glorified laptop watcher 😒) but we're both much happier being able to share our ideas outside of our own little bubble and our products have benefitted too!

Here's to another year of nomading with my favourite person and to the Maker community continuing to grow with our help! Maybe one day we'll all get to meet each other ✌️ 💛

I'm going to attempt to chronologically list all of the Marvel characters Stan Lee had a hand in creating. Working with his genius collaborators, Stan and his team produced an insane amount of iconic characters in a roughy ten year period. (Thread)

With his younger brother, writer Larry Lieber, and Jack "the King" Kirby:

Groot

Tales to Astonish #13 (November, 1960)

With writer Lieber, and artist Kirby:

Fin Fang Foom

Strange Tales #89 (October, 1961)

With the King, Stan gives birth to the Marvel Universe:

The Fantastic Four

Fantastic Four #1 (November, 1961)

With Jack Kirby:

Mister Fantastic

Fantastic Four #1 (November, 1961)

SRI CHAMPAKADHAMA SWAMY, BANNERGHATTA HILL, BENGALURU (KAR)
#bharatmandir

The shrine is dedicated to Sri Vishnu along with Devi Lakshmi and Bhoodevi.
It is built in an area full of Champaka trees.
As champaka flowers are offered to Sri Vishnu here, it is called Champakadhama.

It is believed that King Janmejaya had sarpa-dosha. To get rid of it, he took bath in the nearby pond. He emerged with a golden body and sarpa-dosha was also removed. The pond is called Swarnamukhi and believed to have curative powers.
Sri Vishnu murti was consecrated by him.

It is also believed that this forest area was the Dandakaranya where Sri Ram stayed in exile.
Some believe that the original shrine was built by Pandavas.
The pushkarni/pond is emptied once a year so that the rock at the bottom (which has image of Anjaneya) may be worshipped.

X : Our executive team is concerned that we need to up our game in order to out innovate Amazon.
Me : Do you map?
X : Yes
Me : Like this?
X : No. What's that?
Me : A map of a tea shop.

X : Why is that a map?
Me : Long story, all to do with how space has meaning. To keep it short, maps help people to focus on user needs, the components involved, to communicate missing components and scenario play ideas like staff becoming robots.

Me : They’re also good for measuring and managing capital flow, making investment decisions, removing bias and getting rid of duplication.
X : I don’t see how that helps with innovation?

Me : Well, innovation is a tricky word because we use it to describe many things. If you’re talking about differentiation with the adjacent unexplored then you’re experimenting in the “uncharted” space e.g. immortality with magic provided by the special custom built kettle.

X : That sounds like nonsense.
Me : A lot of what people think will be the next great innovation is nonsense. Actually, most of it is. That’s the nature of the uncharted space, it’s experimental, high risk and generally results in failure.
X : We need more reliable innovation.

Important paper from Google on large batch optimization. They do impressively careful experiments measuring # iterations needed to achieve target validation error at various batch sizes. The main "surprise" is the lack of surprises. [thread]

https://t.co/7QIx5CFdfJ

The paper is a good example of lots of elements of good experimental design. They validate their metric by showing lots of variants give consistent results. They tune hyperparamters separately for each condition, check that optimum isn't at the endpoints, and measure sensitivity.

They have separate experiments where the hold fixed # iterations and # epochs, which (as they explain) measure very different things. They avoid confounds, such as batch norm's artificial dependence between batch size and regularization strength.

When the experiments are done carefully enough, the results are remarkably consistent between different datasets and architectures. Qualitatively, MNIST behaves just like ImageNet.

Importantly, they don't find any evidence for a "sharp/flat optima" effect whereby better optimization leads to worse final results. They have a good discussion of experimental artifacts/confounds in past papers where such effects were reported.