eob

Taiwan resident here. Just wanted to include that “massive outbreak” so far means a few hundred cases nationally.

That’s bad, it might get a lot worse, and we’re all nervous — but just wanted to add the context because “massive” can mean a lot of things with orders of magnitude difference.

Related to cloud-scale ANN: if anyone is interested in testing an embedding service that helps programmers scale NLP operations without the infra hassle, please reach out! (edward.benson@gmail.com)

This thread has been a wealth of information. I’ve had great success with Annoy so far; eager to kick the tires on Vald.

I’m constantly blown away at how high quality SpaCy is with such a small team. Everything: from API design, to documentation, to education, to tool ecosystem.

Congrats on the (preview) release and thank you for the great software. Exciting to see SpaCy pull transformers into the mix!

You might enjoy reading about Taiwan’s experiments in this direction [1]. They’ve used an online system to craft 26 pieces of legislation now with public participation. Now quite full-on “governance by git” but a very cool step in that direction.

Pretty sure it’s one of Audrey Tang’s projects [2]. She’s a member of the cabinet who is also an all-around hacker.

[1] https://congress.crowd.law/case-vtaiwan.html [2] https://en.wikipedia.org/wiki/Audrey_Tang

Any chance you could elaborate a bit? This sounds interesting but I'm not in the know enough to read between the lines.

You were trying to see if someone was sniffing the documents uploaded (and confirmed they were).... or you realized you could use them as a vector though which others would post your materials on websites elsewhere (and they did)?

Just wanted to say that both Dropkiq and website to market it look fantastic. Really nice work.

If you're ever interested in licensing a version to work with custom grammars (not liquid), I'd be interested! (email: ted AT instabase.com)

To the folks reaching immediately for the “revolving door” observations:

DARPA is one instance where churn is a feature, not a bug. Or if not a feature, at least an intentional design choice rather than an accident of corruption.

Various rules about the way the USG works make it really challenging to innovate in the way private industry can. DARPA isn’t perfect, but it’s a pretty good effort to engineer an ecosystem within the USG that optimizes for “wild idea pursuit” rather than “Spec’ed hourly labor” (which governs the rest of the defense complex). Part of that ecosystem is regular rotation of leadership to bring in fresh ideas and agendas. Another part of that ecosystem is the acknowledgement that you have to spend money on 99 bad ideas before you find the one nugget of gold. Basically, the same observation that governs VC portfolio math but the DARPA is interested in capability unicorns rather than economic unicorns.

When I worked in DC, the joke was that DARPA was the only agency in town that would pull your funding for SUCCEEDING. The reason being that if you succeeded, it obviously meant they had set their ambitions too low and needed to re-frame the work to be impossible before funding it any further.

If anyone on the Polynote team is lurking: curious if this is a successor to the great work done by NTeract, which you funded (thanks!)

That project experimented with a lot of interesting themes I see echoed here.

There are several preschools like this in the Bay Area. My son has been going to one for three years in the Presidio. They show up rain or shine (but seek shelter in high wind or forest fire smoke).

It’s been an absolutely wonderful experience.

I was a PhD student at MIT CSAIL a few years ago and recall an incident where a visiting scholar from a corporation who had donated money had to be asked to leave because he essentially scraped every piece of internal data and code in the lab and sent it back home.

What was amazing to me was that he was allowed to return a few months later with the explanation “it was just a cultural misunderstanding”

I want to be careful not to point fingers at a particular country or culture, but that experience taught me that a grey-area version of corporate espionage is _very real_ in academia — an environment that operates with all guards down and no defenses against it.

[Edited to fix tappos]

I just wanted to chime in to thank Chariot for the service they provided.

In my experience, few startups literally make your entire day better. I’ve been using Chariot to commute across SF for about two years and it has given me more time at home with my wife and son, better ability to predict when I’ll arrive at work, and a way to use my commute as a quiet space for the reading I could never quite manage on a crowded bus.

So thanks to the whole team. Very sorry that the numbers didn’t quite work out at the end of the day, as it often turns out in the startup game.

SEEKING FREELANCER FOR HIRE | SF or Remote | React

Instabase is a fast-growing company that builds a platform for data processing. We have a growing set of backend capabilities and would like help building front-end interfaces for these capabilities.

If you are a React developer who wants to get involved in the machine learning and data science space from the UI/UX perspective, this is a fantastic opportunity.

We have a strong preference for contract-to-hire in the SF or NYC region, but are willing to consider pure contracting for the right person.

Contact: ted@instabase.com

Instabase | Machine Learning Engineer | San Francisco and NYC | careers@instabase.com

About Instabase

Founded in 2015, Instabase is a cross-cloud operating system with a powerful suite of data processing applications. We have taken $23M Series A funding from Andreessen Horowitz, Greylock, and NEA, and we have strong customer and revenue growth. Instabase is used in universities (Stanford, Columbia) and large financial institutions in Asia and the US, both as a platform for collaboration and a tool to automate human-intensive workflows. We are a small team, ~15 with 90% engineers, scaling rapidly driven by customer demand. We do core R&D, product management, sales, customer success --- the whole product experience!

Job Description

We take the latest techniques from research labs around the world, evaluate them against each other, and adapt them into products for use in industry. Living at at that edge of research and production is the “secret ingredient” that powers many of our capabilities as a platform.

As a machine learning engineer, you will be responsible for writing and maintaining Instabase's machine learning stack. This role includes developing specific models that enable new apps, designing general-purpose machine learning infrastructure for customer use, and helping design end-user apps that make machine learning capabilities more accessible. We have the benefit of close customer relationships and fantastic dataset access to inform our work: what you do at Instabase makes an immediate impact on our customers’ capabilities.

We look for people who have a bias toward action, who enjoy finding patterns amid chaos, and who are capable of driving a project from whiteboard sketches to completion.

Working in this area requires knowledge in one or more of the following:

- Languages: Python, Go, C++

- Frameworks: OpenCV, NumPy, Tensorflow, Keras, SpaCy

- Mindset: Strong ability to translate ML fundamentals into applications

Unsure if this is related, but a few weeks back AWS paravirtualized instances were experiencing extreme clock skew and suffering frequent restarts. Our clocks were skewing about 8 seconds per hour, and the instances were getting swapped out at least once per day, sometimes more.

Switching to a HVM instance type resolved the problems.

It seems to me the reason for insisting on a verbose declaritivization of everything is obvious: it guarantees you can build run/traintime environments which scale your model automatically.

Google’s mindset isn’t “train this model to multiply by three”. It’s “train this model on a 1% sample of search traffic over the last year.” That’s reflected in the design choices of tensorflow.

For a cool writeup of how to automate an Amazon affiliate site (like this one) that scrapes Reddit for links, check out this post: https://blog.cloudstitch.com/shut-up-and-make-me-money-96c32...

I have long thought the PhD experience is kin to historical tales from Zen monasteries. "Enlightenment" comes not when you pass the master's test, but when you realize you've gotten what you came for and simply tell the master you're leaving.

I allocated and enforced a timeline of only 3 months for dissertation writing -- to the frustration of my committee -- precisely because I had a sense this was how the game worked.

When alien archeologists study the third rock from Sol, they will find no life save for IRC C&C bots chattering back and forth. Feels like the end to an interesting sci-fi novel.

As a former ML researcher, I read your quoted text and came away with the exact opposite conclusion. There is an incredible need as a community to work toward reproducible results.

Pull a researcher aside, and I bet you every one will have at least one experience of trying to reproduce some reported--sometimes even lauded--result but being unable to.

The good news is this kind of cultural fix is really invariant of compute platform or modeling framework. It's just nice that MSFT appears to be making it a priority.

I wasn't able to figure out how to use the app, but I did want to drop in to say that I like how you've placed the "Upgrade to enable" notices in cells past a limited range.

Cool idea. We've built a spreadsheet-powered web platform (Cloudstitch - YC S15) that lets you publish web forms and widgets that rely on spreadsheets as a backend.

Several folks use us just for the spreadsheet API -- as this service seems oriented. We provide a simple, common API that can talk to both GSheets and MS Excel 365, support various forms of row-level access control, and Ruby on Rails-like join syntax. Would love to hear feedback.

It'd be super valuable to see case studies of mixed teams avoiding a common problem in which Techies have the upper hand because they are the mediators of whatever system is actually getting built.

Techies have something of an ultimate veto ("it wouldn't make sense to build it that way") that can really impact everything from brainstorming to product design.

Cloudstitch founder here (YC S15).

We wrap Google Sheets and Excel 365 in an API you can use to GET/PUT data along with a number of other goodies:

   - security policies (e.g., row-based auth)
   - file upload support
   - email triggers upon upload
   - a notion of "frozen" releases versus live dev data
     (released but yet undocumented -- email hello@ for details)

Very happy to provide support for anyone looking to use us as an API into their sheets.

VSCode + TypeScript is the perfect foot-in-door for MS to get into the web space.

They're both excellent products by themselves, but also give Microsoft the platform to start dangling turnkey Azure integration in front of developers.

Imagine if the IDE started to offer the ability to configure, deploy, and manage targeted production stacks--no browser / command line hackery required. That would be compelling.

We (Cloudstitch, YC S15) are building remote work into our company DNA right from the start. It's been an educational journey, and we're stronger team because of it.

One thing we've learned is that "diversity" is so much more than just race/gender/ethnicity. Geographic location, even within the US, can be a profound contributor to the diversity of ideas, experiences, and opportunities. Makes you realize a non-remote team is living in a [geographic] bubble by definition.

If you're interested in building a better web stack / CMS that integrates with your office suite, give us a holler at hello@cloudstitch.com :-)

PS: Re open office plans in early-stage startups, I suspect cost savings are a driving factor.

Cloudstitch | San Francisco | Full-time | REMOTE OK

Join a Y Combinator company (S15) building the future of web publishing stacks. Think GitHub Pages, but powered by MS Office & Google Docs. We handle all the hard engineering under the hood, and our users interact with their sites as if they were just spreadsheets and shared doc folders. We're already powering millions of pageviews for thousands of developers, and we have the whole web in our sights.

We're currently hiring our founding engineering team of full-stack (Node) and front-end developers. Great opportunity to have a large impact and generous equity.

https://www.cloudstitch.com/ Email ted@cloudstitch.com and/or visit https://www.cloudstitch.com/hiring

Polymer offers a lot of UI widgets, which are enormous when you glob them together, but the base polymer library is actually incredibly small due to the fact that's it's leveraging the browser's native custom element functionality.

Can't reply to your comment so editing here:

Works natively on IE11+, Edge, Safari 9+, Chrome, Opera, Firefox, and mobile browsers. The polyfill for everyone else (webcomponents-lite.js) is 41kb.

Looks neat, but also like a non-standard implementation of Web Components, which, though slow coming, already have universal browser support with a small shim.

What's the advantage of this over something like Google's Polymer library, which is a toolkit for rapidly creating reactive web components?

Beyond serving hard tech with longer time-to-market requirements, there is a huge opportunity for brain-rich cities like Boston to build out a more robust startup ecosystem to compete with SF. People in Cambridge joke about the one-way plane tickets to the Bay Area. If you're anchored in Cambridge, that means huge spoils for the group that figures out the equation to keep them there.

Well Harvard iLab is a University-wide incubator that has its own building. They offer advising and limited resource support (e.g., a small lab filled with hardware hacking tools), but no funding as far as I'm aware.