HNHacker News
TopNewBestAskShowJobs

spenczar5

4,833 karma · joined July 12, 2012

spencer nelson

firetiger.com, previously twitch/aws, i like woodworking

s@spencerwnelson.com

submissionscomments
spenczar5··on Crush: Glamourous AI coding agent for your favourite terminal
Yes, I think that sounds like Warp.

Personally, I find warp a bit disorienting, but it is indeed a more integrated approach.

spenczar5··on Parsing Protobuf like never before
JITting bytecode VMs really seem fantastic at this. A similar trick works in Avro in Python (https://journal.spencerwnelson.com/entries/avro.html).

I suppose the insight is that schema documents are programs, just in a weird language, and compiled programs are fast.

spenczar5··on Bypassing Google's big anti-adblock update
Uh sir the article is about JavaScript Browser APIS
spenczar5··on Astronomers race to study interstellar interloper
That is very impressive - getting the nongravs is a lot of work.

Anything published on your integrator and its modifications?

One nice feature of ASSIST (from what I remember, its been a while) was that I could add in more perturbers and crank up the gravitational harmonics if I wanted to. It sounds like you support that too at least for perturbers?

spenczar5··on Astronomers race to study interstellar interloper
Cool to see! I spent a few years working on asteroid orbital dynamics too. What integrator are you using? Do you cover the weird stuff like Yarkovsky effects? That gets important for NEO impact risk, which is what I worked on.

Matt Holman's ASSIST (https://github.com/matthewholman/assist) struck me as a breath of fresh air, coming from openorb and its kin.

spenczar5··on MCP: An (Accidentally) Universal Plugin System
honestly, yes - but MCP includes a really simple 'reflection' endpoint to list the capabilities of an API, with human readable docs on methods and types. That is something that gRPC and OpenAPI and friends have supported as an optional extension for ages, but it has largely been a toy. MCP makes it central and maybe that makes all the difference.
spenczar5··on Building Effective AI Agents
(December 2024, which somehow feels an eternity ago)
spenczar5··on GCP Outage
That's fairly typical. You want a human in the loop for decisions like that.
spenczar5··on How I program with agents
The issue is, whats the alternative? People are generally bad at predicting what work will get broad adoption. Carefully elegantly constructing a project that goes nowhere also seems to be a common failure mode; there is a sort of evolutionary pressure towards sloppy projects succeeding because they are cheaper to produce.

This reminds me of classics like "worse is better," for today's age (https://www.dreamsongs.com/RiseOfWorseIsBetter.html)

spenczar5··on A look at Cloudflare's AI-coded OAuth library
Of course, these are awful for a human. But I wonder if they're actually helpful for the LLM when it's reading code. It means each line of behavior is written in two ways: human language and code. Maybe that rosetta stone helps it confidently proceed in understanding, at the cost of tokens.

All speculation, but I'd be curious to see it evaluated - does the LLM do better edits on egregiously commented code?

spenczar5··on New study casts doubt on the likelihood of Milky Way collision with Andromeda
I could have told you in advance it would have been a Finn! Something in Helsinki has been going on for a while in astronomy. They're so much more willing to challenge the norm, and unusually capable with Bayesian statistics and big computation. It's really remarkable.
spenczar5··on Systems Correctness Practices at Amazon Web Services
I want something like this, but I work almost entirely in Go.

Am I correct in thinking that this would require a fork of the main Go implementation, in order to make a deterministic (and controllable) goroutine scheduler and network stack?

Does one also need to run on a particularly predictable OS?

spenczar5··on DuckLake is an integrated data lake and catalog format
Even though it's in your own SQL DB, there's still some sort of layout for the metadata. That's the thing that trino/bigquery/whatever won't understand (yet?).

> They solve query parallelism by allowing you to perform computations on the edge, enabling horizontal scaling the compute layer.

Hmm, I don't understand this one. How do you horizontally scale a query that scans all data to do `select count(*), col from huge_table group by col`, for example? In a traditional map-reduce engine, that turns into parallel execution over chunks of data, which get later merged for the final result. In DuckDB, doesn't that necessarily get done by a single node which has to inspect every row all by itself?

spenczar5··on DuckLake is an integrated data lake and catalog format
I agree with everything you said. I just mean that a single node may be slow when processing those parquet files in a complex aggregation, bottlenecked on network IO or CPU or available memory.

If the thesis here is that most datasets are small, fair enough - but then why use a lake instead of big postgres, yknow?

spenczar5··on Why Cline doesn't index your codebase
What does that mean? In what way does "search and feed" differ from "retrieve and augment?" Those are practically synonyms to me - is it about tool use vs prompt crafting?
spenczar5··on DuckLake is an integrated data lake and catalog format
There is a lot to like here, but once metadata is in the novel Ducklake format, it is hard to picture how you can get good query parallelism, which you need for large datasets. Iceberg already is well supported by lots of heavy-duty query engines and that support is important once you have lots and lots and lots of data.
spenczar5··on I think it's time to give Nix a chance
Something I have always wondered about is how Nix interacts with my editor and language servers.

I use emacs. The emacs LSP mode starts up a language server process for a language like Python or Go. If I use Nix to manage development dependencies like my compiler, linting tools, and even language-specific dependencies, then how do I get the LSP and emacs to use the correct Nix-y set of dependencies? To make things extra complex, note that I often work on multiple interdependent projects concurrently; I imagine that makes things even gnarlier.

Is there a sane way to manage this sort of thing? Do I end up managing a per-project installation of emacs??

spenczar5··on We fell out of love with Next.js and back in love with Ruby on Rails
Yeah, wouldn't want to rewrite the frontend in a new framework. Good thing the SPA frameworks are so stable and solid; when I choose one I will surely be able to use it for a good, oh, 3 to 6 months.
spenczar5··on All clothing is handmade (2022)
> That doesn't seem to jive with the way I live vs the literal log fucking cabin they lived in.

I propose that different goods have different trajectories. Lighting is better and cheaper than 100 years ago. Boots maybe aren't.

spenczar5··on FOSS infrastructure is under attack by AI companies
Thanks for writing about this. Is it clear that this is from crawlers, as opposed to dynamic requests triggered by LLM tools, like Claude Code fetching docs on the fly?
spenczar5··on The School Car Pickup Line Is a National Embarrassment
It is importamt to understand the whole cultural picture of this practice:

1. Neighbors and friends are notified in advance. "If you see my daughter and she looks lost, please help her, she is doing an errand."

2. The destination is usually notified. "My daughter is coming to buy rice, let me know if she isnt there in two hours."

3. This practice is somewhat old-fashioned and rural. Many parents wouldnt consider it today.

Remember, the show was aired to a Japanese audience. If it were universal, it would not be worth televising.

spenczar5··on Sorting algorithms with CUDA
As the other posts have said, this isn't the right algorithm. Onesweep and its kin are cool but intimidating. The magic is easier to understand if one looks into the core algorithm: radix sort.

A very readable explanation is here: https://gpuopen.com/download/publications/Introduction_to_GP...

It turns out that radix sort can be implemented in a way that is very straightforward to parallelize. It's a beautiful and elegant approach and worth knowing about!

spenczar5··on A drill bit that can also drive screws
> It would be great if every decent sized box of screws came with a drill/screw bit appropriately sized for pre-drill and screw

This is very common. GRK, Spax, Axis (a local brand to the Seattle area), and many more. Yes, they are better!

spenczar5··on AstroForge selects target for “high risk, seat of the pants” asteroid mission
I have directly worked with one of these companies on asteroid discovery. Even they would say it does not cash out for decades. But the premise is that as costs improve it becomes fabulously lucrative.

These companies are very, very risky. If they weren’t, you would see more competition!

spenczar5··on AstroForge selects target for “high risk, seat of the pants” asteroid mission
The usually cited mineral is iridium. It is exceptionally rare on earth (ninth rarest stable element!) but abundant in asteroids and very useful in alloys and many applications.
spenczar5··on Astronomers delete asteroid because it turned out to be Tesla Roadster
Ooh, I know this one! I worked on historical asteroid observations categorization and precovery for a few years.

The MPC has many observations of unidentified moving object candidates. These are called “tracklets” and come from pairs of observations of the same patch of the sky by the same observatory, separated (typically) by a few minutes.

The “isolated tracklet file”, or ITF, is a catalog of all of these unidentified moving objects.

When an identification is made and submitted to the MPC, the MPC back-projects the orbit and checks the ITF for any past observations which might have actually been of this newly identified object.

Then, the designation’s timestamp is of the first matching observation. So in this case the ITF had an observation from back in 2018.

Occasionally, two “objects” turn out to be the same actual physical object, but we learn late. In this case, the MPC does maintain a list of “aliases” of the object, so you might get that “aka” list. But that is not quite what happened here.

spenczar5··on Building a Medieval Castle from Scratch
An interesting case of that is the stave church builders of Norway. Stave churches are marvelous woodworking achievements, and too many of them sprung up too quickly in the historical record. The current theory is that there were traveling bands of master craftsmen who would start the project, do the hardest parts, and teach the local peasants how to finish the job. Then they would move on to the next parish.
spenczar5··on Building a Medieval Castle from Scratch
We can overly fetishize specialization in the past, too, though. Those 6,000 workers were aiming for speed. Amateurs tend to aim for perfect quality.

Sometimes when you look at old things, it can be shocking how “sloppy” the workmanship is, but you have to realize that they were not trying to get perfect paint jobs or smooth castings. They were trying to make money.

spenczar5··on No Calls
> The economy is built on grifting, at this point

I agreed until here. Obviously, lying isn't the only way to make money. I make furniture and fix windows in old houses for a living. Am I grifting?

When you stretch into hyperbole, you lose the ability to convince people in the middle.

spenczar5··on One of my papers got declined today
“do I want to suffer such rejection humiliation? Not really”

The point of Terence Tao’s original post is that you just cannot think of rejection as humiliation. Rejection is not a catastrophe.

← PreviousPage 2 of 24Next →