News for hackers. Stuff that matters.

dleeftink

Re:Neural Networks Are Already Clean on Thursday October 15, 2026 @08:44PM (#50112271)
Why do we want to reduce 'the neuron' and encapsulate it? As a potentially interchangeable feature?

Even when similar weight activations arise across learning regimen it seems we will miss a whole range of interpretations if we just go off 'the text'. The outside context in which a neuron arises is equally, if not more, telling; its genesis cannot (yet?) be quite captured by the weight itself.

The network may reveal its own context and history, but what can ultimately be encoded in its weights that originated outside the network?

You could likely cluster individuals by drawn features; might not tells us much about geographic origin but may reveal interesting differences/levels in geographic aptitude, and which clusters correspond best to what land borders/contintents.
Re:Why is Hacker News like that? on Wednesday September 23, 2026 @12:42PM (#49818805)
> necessary algorithmic check

Are there any stats available on the weighting ratios such as flags to upvotes, likes and comments per post after flagging, average time to frontpage to flag?

(G summarises the following[0], but a definitive source would clear up what the necessary algorithmic checks are).

[0]: https://share.google/aimode/cs9cKsvGMNRHs8OG1

Re:Why is Hacker News like that? on Wednesday September 23, 2026 @12:26PM (#49818535)
Wouldn't excessive/above average flagging rather indicate a post is controversial/counter narrative enough to warrant longer front page exposure? It certainly seems to have gotten the people going in the short time it was up, and it's not a secret that early vocal users can be quite generous with the flag button.
Re:Every Novel Is Boring–Until It Isn't on Monday September 07, 2026 @08:42PM (#49604420)
Flow is a state of privilege, too, yet (thru reading) one most easily cultivated. Take control of your time expenditure!
Re:Tangle – Visual ML Pipeline Editor on Wednesday September 02, 2026 @01:33PM (#49539593)
Seems like a suitable candidate if you ever get tired of KNIME[0].

[0]: https://www.knime.com/why-visual-workflows

Re:Keenable SELECT: an agent that searches the web in SQL on Tuesday September 01, 2026 @12:30PM (#49524285)
Echoes of a semantic Web that never was! Showing the prompt trajectories is also a nice touch.
Re:Ambient CSS v3 – Blender meets CSS on Tuesday September 01, 2026 @12:25PM (#49524200)
Remember motiongerm and the whole sleek 60fps flash design thingamabob? Bespoke interfaces everywhere
There is of course, transformative use, and the slew of software packages we used to depend on without acknowledging the original maintainers on every commit.
They say to throw away the first page. It needs to hurt a bit as we go.

For 'context priming' I'd like to see a tool that supports this. Not just alts and variations, but an assistant or companion that challenges what's already on the page, and sometimes doesn't even look back at what's been thrown out.

Forgetting can be just as effective as capturing.

Beyond imaging, a related paper I came across goes in-depth on identifying the interstellar nodes/constellations to furnish an actual network mesh:

> [ABSTRACT] An interstellar communication network benefits from relay nodes placed in the gravitational lenses of stars. The signal gains are of order 10^9 with optimal alignment, allowing for GBits connections at kW power levels with meter-sized probes over parsec distances. If such a network exists, there might be a node in our solar system: where is it?

In: Hippke, M. (2021). Interstellar communication network (III). Locating deep space nodes. arXiv preprint arXiv:2104.09564[0].

[0]: https://arxiv.org/pdf/2104.09564

Just as stating the wrong thing would be the quickest way to elicit a response in the past, so too can 'dumb' questions prime the context for more complex queries.

I am all for this strategy, and revisiting my list is equal parts fun and conducive to long-term recall.

Software has a dependency contract that is quite unique. We can't imagine the shelflife beyond the Bazaar or Cathedral. They seem ever lasting.

The myth of impossibility remains, because we can't imagine what it means to forever occupy those places, even when others rise past them.

Alternative to resolving an image via Einstein ring lensing?
I think social graphs are often pre-disposed to capturing strong interactions only—an email chain, a trace-able back and forth (just as this), friend degree networks—not the off-hand conversations you overhead on the coffee corner, a tidbit you saw on a documentary awaiting a doctor's appointment, a faraway poster hanging on a museum wall.. you get the idea. Weak ties are everywhere! (Just not always recorded)
The weak ties were strong all along.
Re:Computation as a Universal and Fundamental Concept on Friday July 10, 2026 @04:58PM (#48865166)
As far as I understand it, the storage is the space, or rather, spacetime emerges from a system's informational capacity and the degrees of freedom along which information can spread. On the balance all information is retained but undergoes various phase transitions (the 'computations').
Re:Real-time map of Great Britain's rail network on Monday July 06, 2026 @12:25PM (#48806871)
There are two very lonely trains departing from Stranraer and Kyle of Lochalsh
> it cannot have any ownership rights

Actor network theory sheds a tear. And more promptly: ownership rights are enacted via tools. Whether agentful or non-agentic, listing them would in the very least be similar to disclosing funding bodies or agencies.

Re:Bring Back Crappy Forums on Thursday July 02, 2026 @12:31AM (#48756592)
Depends on what we value, I suppose; a depth-first style that surfaces isolated chains or a breadth-first style that surfaces interleaved replies.
Re:Claude Science on Tuesday June 30, 2026 @02:49PM (#48737478)
Some authors may even choose to leave syntactical errors as a tell for those self-authored passages; long-term, some interesting language drifts may come of it.
Re:A Herculaneum scroll has been read for the first time on Thursday June 25, 2026 @02:38PM (#48677530)
Just as with redacted documents (consistently blocked terms) or bad OCR jobs (wrong or missing characters), even if only a certain percentage comes out unmangled it is more readable than having no data at all.

A stable base corpus and some dynamic programming will allow you to clean up the remainder[0].

[0]: http://stackoverflow.com/a/11642687/2449774

Re:Pull request limits are cutting down the noise on Wednesday June 24, 2026 @04:02PM (#48664939)
Hence the cooldown period? I think the mechanism proposed here should be perfectly fine for targeted PRs, while mitigating those that sit above baseline.
Re:Modos Color Monitor Pushes E-Paper Displays Further on Thursday June 18, 2026 @11:39AM (#48587084)
> Don’t make yourself regret the things you didn’t do

Nothing to add, but it bears repeating. A shimmer of indie tech resilience

Re:Am I Unc? on Friday June 05, 2026 @01:41PM (#48415815)
The real unc is non-trendy so therefore new balances are non-unc, which makes them unc
Re:"They're made out of weights" on Thursday June 04, 2026 @04:14AM (#48395693)
Re:Pluto.jl 1.0 release – reactive notebook for Julia on Wednesday June 03, 2026 @08:14AM (#48382968)
Looking forward to using this[0] with Pluto instead of repl! (They just look so cool)

[0]: https://github.com/JuliaPlots/UnicodePlots.jl

Single solutions/solutionistic approaches will likely be incompatible with either goal; consumer needs are always changing and collection capabilities expanding. Data scope and retention also need not be counter to consumer wants, and in the very least requires a mechanism that allows consumers to 'dial in' their preferences rather than wholesale accepting/rejecting terms of usage (i.e. a gradient instead of a binary).

I've yet to encounter a service that has implemented this successfully.

Re:My two-part desk setup on Saturday May 23, 2026 @03:42PM (#48250713)
They'll only get you so far
Re:My two-part desk setup on Saturday May 23, 2026 @02:40PM (#48250100)
I saved my desk from curb side collection. My chair idem. My laptop battery died two years ago so my desk cannot be too far away from a wall socket.

Maybe one day I could face my desk away from a wall.

Older »