HNHacker News
TopNewBestAskShowJobs

georgewfraser

2,609 karma · joined October 28, 2012

fivetran.com

twitter.com/frasergeorgew

george at fivetran dot com

submissionscomments
georgewfraser··on How to pick a good monitor for software development
I’ve bought two monitors that I really loved: an Apple Cinema Display 27” for $2000 in 2003, which was a crazy amount of money for a monitor at the time. I used it for 12 years. And an Apple Pro display with all the options for $7k last year, which seemed a little crazy but it’s just unbelievably good, like looking through a window, and I won’t be surprised if I use it for 10 years. If you’ve got the money, a really good monitor is worth a lot.
georgewfraser··on Building a modern home in the woods
It’s at least $1,000/sqft all-in. Construction costs have gone up a lot in recent years, and architecturally designed houses like this are expensive to design and build.
georgewfraser··on Building a modern home in the woods
It’s too bad they didn’t share their final costs, there’s very little public info on how much custom builds like this really cost. I think people are a little embarrassed by how much they spent and so they don’t want to reveal it.
georgewfraser··on Amputated fingertips sometimes grow back (2013)
This is one of my favorite weird biological facts. It actually happened to a friend of mine. It shows that the programming for regeneration is in there, we just need to activate it.
georgewfraser··on 2021 Tesla Model Y review: Nearly great, critically flawed
Doh, how did I miss that? Now I understand: the phantom breaking is all on non-highways. I only use autopilot on highways and highway-like roads.
georgewfraser··on 2021 Tesla Model Y review: Nearly great, critically flawed
I don’t know how to square this review with my own experience of owning a Y for a year. The autopilot is nearly perfect, noticeably better than a human driver in stop and go traffic on the highway. Given the history of downright dishonest reviews by old school car people with an irrational hatred of Tesla, this review really should have been accompanied by video evidence of phantom braking.
georgewfraser··on YAML: It's Time to Move On
No recursion and no lambda.
georgewfraser··on YAML: It's Time to Move On
A radically different alternative with a lot going for it is Starlark: https://github.com/bazelbuild/starlark

It’s a deterministic subset of Python. This means that if you have complex or repetitive configurations, you can use loops and functions to structure them. But it’s impossible to write an infinite loop or recursion.

georgewfraser··on Cortical Column Networks
The fact that neurons on top of each other have similar firing properties is very interesting, and is indeed a uniform organizing principle of the cortex. Calling this “minicolumns” is very misleading because it implies some level of horizontal organizational structure. That is not a universal organizational principle of the brain—it exists in certain areas, in certain species. It’s an accident of history that a lot of the early neuroscience in mammals was done in primary visual cortex in macaques and primary sensory cortex of rats and mice, which do work like this (though at very different spatial scales). This led to the hypothesis of the “cortical column” as a universal, intermediate level of structure in the cortex, but when you look at other areas they just don’t work like that. However this idea is so attractive that it just won’t die.

Meanwhile, cortical column fans (including neuroscientists!) engage in a motte-and-bailey argument, where the bailey is “the cortex is divided into columns that represent an intermediate level of modularity” and the motte is “neurons on top of each other usually have similar firing properties”.

georgewfraser··on Cortical Column Networks
Disclaimer: have personally recorded from whisker barrels, the structure that inspired Vernon Mountcastle to posit the idea of the cortical column. He’s my neuro-great-grandfather according to neurotree: https://neurotree.org/neurotree/tree.php?pid=154706&fontsize...

The idea of cortical columns isn’t like the 6-layer-cortex. The cortex just isn’t organized into little column-shaped modules. It’s true that if you record two neurons right on top of each other, they’ll have a similar receptive field/preferred direction/…. BUT there is no consistent horizontal organization of neurons into columns. There’s whisker barrels, which reflect the way rat and mouse whiskers work, and there’s ocular dominance columns, which are a totally different thing at a different spatial scale. But there’s no global organizing principle that should be called “cortical columns”. It’s a lovely idea that would be great if it were true—-it would be so much easier to understand the brain if there were a common unit of organization larger than a neuron but smaller than a brain! But it’s not so.

georgewfraser··on Cortical Column Networks
The cortical column was a hypothesis about a general organizational principle for the cortex, which turned out to be more or less false but which has lived on as a kind of “zombie science” because too many people built their careers around it and we can’t admit that it isn’t a real thing until they all retire. The thing that actually exists is “neurons above and below each other tend to respond to the same stimulus/action in the real world.” Jeff Hawkins, being a smart neuroscience-amateur, seems to have never realized that cortical columns are this nice idea that was found to be wrong a long time ago, but neuroscientists don’t say that out of politeness to the profs who built their careers on the concept. For more detail: https://pubmed.ncbi.nlm.nih.gov/15937015/
georgewfraser··on A Multi-Billion Dollar Real Estate Project Is Rising on Native Land in Vancouver
It’s fantastic that they’re bypassing the NIMBYs and building high density, but it’s terrible that they’re building the 10,000th attempt at modern architecture’s worst idea, “towers in the park”, which fails every time it is tried.
georgewfraser··on The Rise and Fall of the OLAP Cube (2020)
A data mart is a logical organization of data to help humans understand the schema. What I am describing is a physical optimization, extremely similar to what an OLAP cube would do, but implemented on top of a SQL data warehouse. It’s an orthogonal concept to a data mart.
georgewfraser··on The Rise and Fall of the OLAP Cube (2020)
It’s much smaller than the original table. If you compute lots of these, then voila, you have an OLAP cube.
georgewfraser··on The Rise and Fall of the OLAP Cube (2020)
The OLAP cube needs to come back, as an optimization of a data-warehouse-centric workflow. If you are routinely running queries like:

    SELECT dimension, measure
    FROM table
    WHERE filter = ?
    GROUP BY 1
You can save a lot of compute time by creating a materialized view [1] of:

    SELECT dimension, filter, measure
    FROM table
    GROUP BY 1, 2
and the query optimizer will even automatically redirect your query for you! At this point, the main thing we need is for BI tools to take advantage of this capability in the background.

[1] https://docs.snowflake.com/en/user-guide/views-materialized....

georgewfraser··on Exploring performance differences between Amazon Aurora and vanilla MySQL
I wonder why Aurora shares undo logs between replicas? It’s perfectly possible for the read replicas to each re-create their own copy of the undo logs, and retain those undo logs for different durations based on the different long-running queries on each replica.
georgewfraser··on The Query Your Database Can’t Answer
Anyone thinking about using Confluent as some kind of alternative to a database should read this blog post outlining the myriad correctness problems with ksqlDB: https://scattered-thoughts.net/writing/internal-consistency-...
georgewfraser··on Distributed Cloud Builds for Everyone
sccache[1] is a similar project, which supports remote execution and caching for C, C++, and Rust. Unfortunately the remote-execution mechanism is designed for Mozilla's internal environment and doesn't support cloud backends like Lambda or Google Cloud Build. But the code is well-structured Rust and not too big, adding a cloud backend would be a nice project.

[1] https://github.com/mozilla/sccache

georgewfraser··on The data model behind Notion's flexibility
I was thinking more along the lines of, specific tables for different classes of entity, with a schema that’s as fixed as possible, so you get the most compact and indexable representation.
georgewfraser··on Project Starline: Feel like you're there, together
I have always wanted to try setting up a low-tech version of this, where instead of using a light-field display, you assume that you have only one viewer and you simply render the correct point of view on the screen. This is the same technique used to shoot The Mandalorian [1], but in this case we're only trying to make a "pane of glass" so an ordinary high-res TV will do. With both participants positioned correctly, you will get correct accommodation, vergence, and parallax depth cues---only stereopsis is missing, and stereopsis is much less important than people think. 1/3 of people don't even have good stereo vision! [2]

As long as you're willing to give up stereopsis, and you have only 1 viewer on each side, I think you could accomplish this type of immersive telepresence with an ordinary TV + software.

[1] https://www.youtube.com/watch?v=Ufp8weYYDE8 [2] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4934608/

georgewfraser··on The data model behind Notion's flexibility
Salesforce and JIRA both did something similar: their underlying database schema is very generic, basically keys and values, allowing arbitrary logical schemas to be defined at runtime. Yet in both cases, they ended up not really taking advantage of this flexibility. The logical schema of both systems is a very ordinary relational schema that could have been implemented directly on the database, with much better performance. I wonder if the Notion developers made a serious attempt to build on top of a more conventional structured schema, and found it really was unworkable?
georgewfraser··on Generational changes pressure CEOs to take up political activism
This is a mostly-fake, media phenomenon. Most CEOs and most companies do not engage in political activism, we just tend not to notice people not speaking out. What's described in this article is something the authors (and most journalists) want to happen, but it isn't actually happening at most companies.
georgewfraser··on A skeptic’s take on Neuralink and other consumer neurotech
I did my PhD in this field, I love that Neuralink is doing what they’re doing, but I also agree with this article. Your hands on a keyboard is a fantastic brain computer interface and represent a high bar that an implanted BCI must clear to add value for non-paralyzed people.
georgewfraser··on Opensource.cooking: A simple, bloat-free cooking site
Using the joy of cooking instead of google for finding recipes was a funny revelation for me. The index of a book is so much better than web search for this scenario:

- Only one recipe.

- Vetted to be high quality.

- Gets straight to the point.

The internet is amazing but for a lot of domains books are way better.

georgewfraser··on Exodus: Easily migrate your JVM code from Maven to Bazel
Bazel guarantees that each build step depends only on its declared inputs, and that each step is deterministic. So if the inputs to a step have not changed, Bazel can skip that step and use a cached copy of the result of that step. If you break up a large multi module build into a tree of small steps, Bazel can perform an incremental build after a small change in a few minutes, even though running the entire build from scratch would take hours.
georgewfraser··on Tesla owners asking what happens if 'full self driving' isn't real
I bought a model Y with FSD a few months ago. It’s fantastic. “Navigate on autopilot” is game changing for highway driving. It’s 80% a useful as a completely self driving car.
georgewfraser··on Build your own “data lake” for reporting purposes
Typo :)
georgewfraser··on Build your own “data lake” for reporting purposes
Replicating a Postgres instance to a data warehouse takes ~1 hour to set up and $500/month, depending on the database size. The cost of the data warehouse is similar. If you can’t afford that cost, you’re not going to be able to afford an analytics team to make meaningful use of that data. In that scenario, you’re better off using Google analytics + the occasional ad hoc query of your production DB (which you should only have one of, you are at way too small scale to have multiple DBs).

Federated data systems, like what is described in this post, are a kind of “nerd trap.” They’re fascinating and people keep building them, but they have little real world utility. Everyone is either big enough where it’s worth it to build a real data warehouse, or small enough that they shouldn’t be messing around with complex data infrastructure.

georgewfraser··on Build your own “data lake” for reporting purposes
This is a needlessly complex solution. You will get better performance, with simpler maintenance, by replicating everything into an appropriate analytical database (Snowflake and BugQuery are both good choices). Setting up multiple Postgres database and linking them together with foreign data wrappers is interesting to blog about, but it’s an extremely roundabout way to solve this problem.
georgewfraser··on Stay-at-home policy is a case of exception fallacy
The entire category of "is X associated with COVID" research is a recipe for p-hacking, because you're looking at relationships between slowly changing variables, with all sorts of confounds, and there are so many ways to do multiple regression/synthetic controls/and so on. It ends up telling you as much about the priors of the researchers as anything else. My favorite example of this is the research group that found that the Sturgis motorcycle rally increased COVID:

https://onlinelibrary.wiley.com/doi/full/10.1002/soej.12475

But the same research group found that BLM protests did not:

https://www.nber.org/papers/w27408

It's possible that both are true, but it's a bit suspicious that these findings are so well-aligned with the probable political associations of the authors.

← PreviousPage 2 of 11Next →