Comparing the state of programmable money to “mainstream” financial services

Source: https://matcha.xyz/

The concept of “programmable money” made possible by the blockchain has gained a substantial amount of traction this year, but this is just the beginning.

With about $13B assets in decentralized finance, the space is evolving at an extremely rapid pace. However, it is still in a nascent phase with the potential of bringing new paradigms to financial services. To create a mapping from mainstream financial services to defi, we will use the fintech space as a proxy.

Fintech has unbundled banks and almost every other traditional financial institution. Moreover, those institutions otherwise known as “traditional” are also going out…


Payments Systems in the US by Carol Coye Benson, Scott Loftesness and Russ Jones was among the most informative books I read on the financial services industry. Below are some key snippets I’d like to retain and share.

A well-savored book :)

Economic Models for Payment Systems

The payments industry is different from other processing industries in terms of the value of money being transferred through the system. Providers who realize revenue related to gross value of the payment transaction (the “amount”) are more likely to have profitable businesses than those who realize revenue simply on a fee-per-transaction basis.

The Network Business (i.e. Mastercard, Visa)

  • Handles transaction switching among banks participating in network
  • Net settlement…


Delighting customers through banking data feeds

Brex’s diverse customers use a variety of software tools to run and grow their businesses. About half of all our customers use some form of ERP (Enterprise Resource Planning) application and expect Brex data to seamlessly integrate with their existing productivity tools.

In late 2019, we identified two sizable opportunities:

  1. Customers demand robust integrations with their existing workforce productivity tools.
  2. Brex’s channel partners like accounting apps can provide a better customer experience by directly accessing customers’ transactions and statement data, at their permission.

We built the Accounting API to enable ERPs to pull data from Brex through a direct transaction…


人与人之间的差别是什么?

张一鸣(字节跳动)

“张一鸣比别人聪明吗?不是。他后来也说:同期毕业的同学,比他聪明的人很多,比他能干的也多。

张一鸣有多勤奋?

2005年,张一鸣大学毕业,开始北漂。仅用2年时候,就从普通程序员做到技术高管。 管理四五十人的团队,负责所有后端技术,同时也负责产品相关的工作。

有人问张一鸣:为什么你在第一份工作就成长很快?是不是你在那个公司表现特别突出?

然而在酷讯工作的2年里,多的是清华、北大、斯坦佛的计算机专业硕士、博士。 张一鸣的技术不是多么出色,更没有什么牛逼的经验。

如果说表现突出,那张一鸣最突出的,可能就是勤奋。”

来源:https://new.qq.com/omn/20191220/20191220A0LF1A00.html

王兴(美团)

“从校内网到美团网,王兴接连引领创业风潮,为什么每一次都能领先一步?每一次都能捕捉到互联网浪潮的浪花?很多人归功于聪明,但王兴自称也不是天才,他的聪明更多的是经过长久的自我训练而培养起的一套行之有效的思维方式。这里,他推荐了一本书《异类》,这本书中说道一些人能成功,是因为泡在这个行业中超过一万个小时,这一万个小时全情拥抱、全心投入、全力以赴,才能对这个行业的趋势、细节更敏感,触类旁通。比起聪明,好学才是最大的特点。而王兴所表现的聪明是长期学习之后形成的结果。”

来源:https://www.jianshu.com/p/a020b767ad1e

李彦宏(百度)

“我没有觉得我比别人聪明多少,我只是觉得我比大多数人更执着。我觉得很多时候人作出的事情另别人觉得很厉害,更多的是由于他的坚持导致的,而不是他先天的聪明。人和人之间的聪明程度真的没有差那么远。更多的是他的情商,他自己的理想和学习的意愿,导致他作出别人做不出来的事情。”


Reading notes

Nothing positive ever comes from a currency war.

Brazilian finance minister Guido Mantega flatly declared in late September 2010 that a new currency war had begun.

At the heart of every currency war is a paradox: currency wars are fought internationally but they are driven by domestic distress.

They begin due to insufficient internal growth. High unemployment, low or declining growth, weak banking sector and deteriorating public finances.

Difficult to generate growth in purely internal means. Promotion of exports through devalued currency becomes the engine of last resort.

GDP is characterized by Consumption, Investment, Government spending, Net exports (exports minus imports)

GDP = C + I + G…


Takeaways from my first job out of college

Sunny evening in Apple Park

After almost 2 years, I wrapped up my time as a software engineer in the Siri Search team within AI/ML at Apple. I’m lucky to have joined an action-packed group within a company I’ve admired since childhood — one that has shaped humanity and became the most beloved brand on the planet. Goodbyes were bittersweet, but I take solace in knowing that Apple will always be part of my identity, like a package/dependency in the software within me.

Fun fact: The Breakout List (known for featuring hot startups) included Apple Siri in its list of “High Potential and High Growth…


Notes from Practical Deep Learning for Coders 2019 Lesson 7 (Part 1)

Other lessons: Lesson 1 / Lesson 2 / Lesson 3 / Lesson 4 / Lesson 5 / Lesson 6

Quick links: Fast.ai course page / Lecture / Jupyter Notebooks

MNIST CNN

course-v3/nbs/dl1/lesson7-resnet-mnist.ipynb

By now, we should be pretty familiar with the process of loading in image data and creating a DataBlock ( likeImageList):

  1. Specify the path of the image data
  2. Load in the data
# convert mode specified is for black/whiteLoad in the data
il = ImageList.from_folder(path, convert_mode='L')

Inside an items list il is the image you gave it, so you can index into the list and view the image content.


Notes from Practical Deep Learning for Coders 2019 Lesson 6 (Part 1)

Other lessons: Lesson 1 / Lesson 2 / Lesson 3 / Lesson 4 / Lesson 5 / Lesson 7

Quick links: Fast.ai course page / Lecture / Jupyter Notebooks

Topics

Techniques for avoiding overfitting

  • Dropout: remove activations at random during training in order to regularize the model
  • Data augmentation: modify model inputs during training in order to effectively increase data size
  • Batch normalization: adjust the parameterization of a model in order to make the loss surface smoother.

Convolutions

Lecture announcement: platform.ai allows you to train models using images. You can use this as a tool to train models on unlabelled data.

Regularization


Notes from Practical Deep Learning for Coders 2019 Lesson 5 (Part 1)

Other lessons: Lesson 1 / Lesson 2 / Lesson 3 / Lesson 4 / Lesson 6 / Lesson 7

Quick links: Fast.ai course page / Lecture / Jupyter Notebooks

Reviewing some concepts from last lecture — remember that activation functions are element-wise. The function is applied to each element in the input.

So if the input to an activation function is a 20-element long vector, the output will be of the same size. ReLU is the main one we’ve looked at.

Universal Approximation Theorem: If you have big enough matrices, it can solve any arbitrarily complex mathematical function to any…


Notes from Practical Deep Learning for Coders 2019 Lesson 4 (Part 1)

Other lessons: Lesson 1 / Lesson 2 / Lesson 3 / Lesson 5 / Lesson 6 / Lesson 7

Quick links: Fast.ai course page / Lecture / Jupyter Notebooks

We continue to look at NLP.

NLP

25,000 movie reviews in the IMDB dataset. It’s not enough information. So the trick is to use transfer learning!

The idea is we use a pre-trained model that has been trained to do something different to what we’re trying to do. …

Julia Wu

Engineering at Brex, prev. Apple, MSFT, Brown CS + Econ. More at juliawu.me

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store