Congratulations to Jacobin for, the day after Lula's defense team filed it's 2nd motion to dismiss due to illegal US involvement in Lava Jato, its timid reference to the US DOJ partnership. This relationship has been public knowledge for over 5 years but better late than never.

Surendra Ananth @surendraananth, you violated my conditional consent. I assented to unprotected intimate relations with you on the STRICT condition that you tell me if you have relations with anyone else. You failed to do so, and I became ill as a direct consequence.

You refused to reveal your own STI results to me, but claimed to others they were negative and alleged I had multiple other partners, slut-shaming me. You gaslit me saying you had told me about your other partners.

You called up @justice_sisters and @SistersinIslam to tell them I was crazy, jilted. You rebuffed attempts to organise a community mediation so we could address matters. You said I was crazy for speaking to your boss and father in an attempt to keep other women safe.

You have a PATTERN OF BEHAVIOUR of misogyny, harassment and sheer disrespect for women. This was sent to a friend of mine who has never expressed interest in you and has a partner.

You think women are playthings, notches on the bedpost. You, @surendraananth, are no human rights advocate.

🛠️ OPEN BUSINESS IDEAS 🛠️

Below a list of business/product ideas I had or read about.

They are worthless if they remain ideas and if you don't overcome the challenges in building them, so feel free to copy / tweak / implement them!
👇

Better: use one and make a MVP during the #24hrstartup challenge!

😻 Product Hunt Time
A clock that displays the time it is @ProductHunt
Also displays what you should do and where you should post, at each specific time during your launch

🚧 IndieCrunch - VC free tech news
Techcrunch but only for bootstrapped companies

🎧 Kickstarter for audiobooks
A lot of awesome books are not available as audio.
Crowdfund the money to buy the audio rights + a voice actor

Your Most PRECIOUS Currencies:

-Energy

-Time

-Attention

-Trust

-Love

-Loyalty

-Respect

Only give these currencies to high value people. Never spend these currencies on low value shitheads eg: trolls, weasels, snakes, naysayers etc.

'Wait, I thought money was the most important currency?'

Nah.

Once you adopt abundance mentality, you realize their is no shortage of money.

Making the dollars your most important currency will have you leading an empty life.

Time to flip your perspective.

👇

▪Energy

When you strip yourself to the core, you are an emotional creature.

Emotions are your internal worlds energy.

Harnessing that energy is crucial for leveraging yourself to obtain whatever you want.

You have a finite amount everyday, so spend it wisely.

▪Time

A second that is lost will never be returned.

You start valuing the hell out of this currency the more you mature.

As the years start adding up, you realize time is precious.

You must always have a scarcity mindset towards time.

Once you do so, you will not be lazy.

▪Attention

You can be here, but not present.

Attention is completely mental.

Giving someone your attention means you are clearing up mental bandwidth to make room for them.

Only give your attention to people who help you grow.

For the negative ones?

Ignore their existence

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.