probdist

Hacker News has incredibly good long term retention of active users (commenters/submitters). I just published a blog post on the subject (https://probdist.com/2020/03/28/the-graying-of-hacker-news-u...).

Lots of users on HN are actively contributing for years after sign up. This to me is a good sign of health. The counter point is of course, the rapid drop off in engagement after the initial month. Where a large number of users each month never return to the site.

The average comment on Hacker News last month was written by a user with 5 years of tenure on the site. Depending on your perspective you could argue this is a good or a bad thing (new users aren't contributing as much as old users, or old users are continuing to contribute!).

OP here: I'll keep checking in on the comments from time to time. If you are after the raw data in BigQuery it is available here: `bigquery-public-data.hacker_news.full`

This is how RStudio and certain other great open source tools get bigcos to pay for licenses. They are allergic to AGPL and therefore will happily pay for alternative licenses for tools their employees demand.

This appears to be an open sourcing of the Databricks Delta technology which is fairly mature and I personally think very valuable.

I'm a former early employee at a KP backed start up. Nothing but respect and good vibes from the folks I met with or presented to from KP.

This is a good move from Slack. Microsoft teams is an extremely high quality product already. Slack has to be wary of this competitive force nipping at their heels. I've seen teams make good penetration into large companies by starting as the alternative to sharepoint for new projects/working groups.

Team's live editing functionality of excel files is really really valuable for team collaboration. Slack closing the gap on O365 integration is important for enterprise sales.

Azure Data Explorer/ Kusto is more of a database that is optimized for the log use case than a service. There is a front end tool and a lot of the use-cases are around log management, but it is database you can do general SQL or KQL things with. Time series is one of the core use-cases for it also but it has less marketing around it.

Seems positioned to compete with Azure Data Explorer (MSFT's log/time series optimized service). I know Azure runs a lot of services on top of Data Explorer (previously called Kusto) I wonder if this is a true internal battle tested product or a me-too offering.

I've seen a broken link from this url parameter already.

Went to a similar event at Harvard some years ago. Was interesting but if you are a HN regular you probably get as much out of blog posts following happenings online as at the event.

YMMV if you are a better net-worker than I am.

Azure has already had a great spark experience available with HDInsight.

I've help sell software-as-a-service to large businesses for a few years and this is a good read. An interesting additional nugget is that when the stakes get big enough customers often are being guided by a consultant (Accenture etc) in crafting things beyond an internal purchasing/procurement/IT mashup. You need to make sure you help the customer and the consultant to get the sale.

Looks pretty neat. Reminds me a bit of Linqpad, https://www.linqpad.net/ which I've also never used.

Why does this support JIT but not AOT for NVIDIA GPUs?

They could have asked for George's middle name and the cost of compliance would be burdensome. It is not the rap but the ride that you can't beat.

There are many of the substantive tweets preserved in the legal record of the case.

Previous (short) discussion: https://news.ycombinator.com/item?id=10954709

How'd you decide what groups you wanted to focus on? Vets, Women, URM, International.

Any other demographic categories you think are important to especially reach out to?

Wow the complaint has some bombshells:

In particular I'd highlight this as pretty crazy: (screenshot from scribd because I don't have a plaintext copy) http://imgur.com/kaFSSzl

One of the Customer Support Staff Members on the Silk Road was a homeland security investigator. Wow.

Oh I thought that these were all going to be street level data from toll booth type things.

I'm missing where these records are present. I basically see plate numbers timestamps and coordinates no other columns on these datasets.

I'm interested in thinking through several examples on this subject.

One example that comes to mind with me is the campaign by many university student groups across the country to encourage endowments to divest from fossil fuel companies. For example: http://www.fossilfreeuc.org/ http://www.fossilfreemit.org/ . The the vocal minority on this subject is very strongly in favor of their opinion (divestment), whereas the larger mass of students mostly doesn't care about the issue or perhaps leans to not divesting but only slightly.

It is interesting to consider the more detailed results of this where people don't have 0 cost acquiescence, for example if people really wanted some non-kosher soda it would still exist in the marketplace.

It is also interesting how this phenomena occurs in voting schemes where you have a fixed number of voting points to distribute over all issues up for consideration. In such systems people can have their minority issues more strongly represented.

This is really nicely produced.

Can someone provide some background for the scene where Mission residents are opposing new luxury condo development? What has been the state of new high density development in SF? Last time I was in SF I didn't see a lot of construction cranes.

I assume the CMU folks didn't break into any one elses nodes. It seems like one of the principles of Tor node operation is that you can run your nodes however you want. I don't think there is any binding agreement or even strong implication that someone operating a Tor node agrees to not be hostile to users.

Perhaps it was unethical by the standards of university research, but I wouldn't be bothered by the government doing this. If the government or a university researcher idled in your irc channel and logged the communications I would think that was okay. I kinda view this the same way.

I'm slightly concerned about the Hadoop/Spark + machine learning ecosystem right now.

There seem to be a lot of technologies and projects being built up on the core foundation of in memory distributed computation. MLlib in Spark, H2O, Apache Mahout/Samsara, probably many others.

My impression is Samsara and SystemML's DML are being designed to supply the primitives you need to build a machine learning model with less awareness of the underlying system model. So this is ostensibly to save the developer of a new algorithm the headache of thinking how their algorithm needs to relate to the distributed computation model like one should consider for developing with H2O or Spark. However, It seems like many common off the shelf algos are already available and implemented indicating that perhaps working with Spark or H2O backends directly is not that hard to produce performant ML codes.

As a computational scientist learning a DSL to accelerate development of an algorithm of mine into production via Spark seems like a potential distraction from just getting better with Spark.

Yes. The link for this story adds less than the average comment on HN in the previous discussion to the overall conversation.

I tend to get conference call setups running via Microsoft Skype for business. The call in number stays the same but the conference code changes for each meeting even if it is a recurring meeting as does the link for joining via web.

Why not just hire a lawyer to help with terms and closing? I'm assuming you have a significant fraction of a house in value here. You'd have a lawyer to help close that transaction.

Indeed. It actually puts UC Berkeley in a pretty horrible position by going public like this. It does not look good for employers to have the same kind of public litany of complaints against an employee.

Some things stand out from a quick skim of the personnel file/included complaint. Among these: he did not practically assign homework or have regular quizzes. This was noted as not being consistent with department norms. It actually _is_ of value to have consistency between multiple sections of introductory classes and as preparation for follow on classes in the same sequences (e.g Calc 1, 2 3 having similar structures in class organization) he seems to have disagreed philosophically with his boss (department chair) on this front. There are some other parts of the files posted that suggest he apparently did not work well with his graduate student instructors (teaching assistants). In my mind those are big enough reasons to let someone go.

He might be a great instructor. It might be a huge mistake for them to let him go, but it certainly doesn't seem like the only available explanations are completely specious or wrongly-motivated reasons. Also missing from submitted link is some info in the complaint that this was attempted to be worked out with a new incoming department chair, as evidenced by this being posted it seems like that didn't work out.

so Bostonography has been a long running attempt to build crowd-sourced definitions of neighborhoods. I expect the interested difference from my maps will be an aggregated report on the data that will come out as has happened on several previous versions of the survey / site.