Hacker News: fastball

New comment by fastball in "Canceling "Hey""

fastball — Mon, 27 Jul 2026 19:12:03 +0000

So let's say hypothetically we kept the number of homeless people stable in Oakland, but made them all white. You think all (most? many?) of the people currently complaining about homelessness would shut up?

New comment by fastball in "Canceling "Hey""

fastball — Mon, 27 Jul 2026 18:19:24 +0000

Homelessness is pretty race neutral tho? Especially in Oakland, which is incredibly race-balanced.

New comment by fastball in "Decker, a platform that builds on the legacy of Hypercard and classic macOS"

fastball — Mon, 27 Jul 2026 02:27:37 +0000

It is actually definitely yes! You can script with JS. Steve Ruiz demonstrates having an LLM do that for you in his intro video to this "offline" version of tldraw: https://www.youtube.com/watch?v=vCyENfxgyYA

New comment by fastball in "Plasma Tunnels Reveal How Dying Satellites Fall to Earth"

fastball — Mon, 27 Jul 2026 01:21:45 +0000

The typical meaning of space junk is stuff orbiting around without the ability to control itself and without de-orbiting, therefore presenting a risk/nuisance to other things in orbit.

Is a Starlink sat that reaches EOL and sets itself on trajectory for rapid de-orbit ever junk by such a definition?

New comment by fastball in "Plasma Tunnels Reveal How Dying Satellites Fall to Earth"

fastball — Mon, 27 Jul 2026 00:35:18 +0000

> However, the rapidly increasing numbers of launches to low Earth orbit now threaten to increase the levels of space junk reentering the atmosphere.

Isn't it not space junk if it is burning up in the atmosphere? Kinda conflating two different concerns here that are somewhat mutually exclusive.

New comment by fastball in "Decker, a platform that builds on the legacy of Hypercard and classic macOS"

fastball — Sun, 26 Jul 2026 19:53:15 +0000

If you want the "agentic future" version of this (sorry), tldraw just relased "tldraw offline"[1] which in some ways has a lot of overlap with HyperCard as well.

[1] https://offline.tldraw.com/

New comment by fastball in "The new rules of context engineering for Claude 5 generation models"

fastball — Sun, 26 Jul 2026 17:11:13 +0000

Is it? There are shills and bots all over the internet, and referring to the model as ChatGPT is probably common for some people, but not for software engineers that actually track and compare the models (at least in my experience).

New comment by fastball in "The new rules of context engineering for Claude 5 generation models"

fastball — Sun, 26 Jul 2026 05:09:23 +0000

The fact you called it ChatGPT instead of Codex makes me suspicious of this entire comment.

New comment by fastball in "JetZero"

fastball — Sun, 26 Jul 2026 04:21:10 +0000

Gimbal the floor?

New comment by fastball in "Inflect-Micro-v2: complete voice in 9.36M parameters"

fastball — Sun, 26 Jul 2026 04:19:01 +0000

With so few parameters, I imagine a voice fine-tune might be readily tractable.

New comment by fastball in "SpaceX Starship Flight 13 livestream [video]"

fastball — Sat, 25 Jul 2026 00:12:49 +0000

Re-light seems to be one of their biggest remaining obstacles in a lot of these flight tests.

New comment by fastball in "GigaToken: ~1000x faster Language model tokenization"

fastball — Wed, 22 Jul 2026 20:20:08 +0000

The problem with the rest of inference is that changes are not trivially correct or incorrect, as they are with the tokenization layer.

New comment by fastball in "GigaToken: ~1000x faster Language model tokenization"

fastball — Wed, 22 Jul 2026 20:18:30 +0000

Tokenization is <0.1% of the inference time for the first token in the same way it is <0.1% for the last.

New comment by fastball in "Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber"

fastball — Wed, 22 Jul 2026 16:19:08 +0000

Claude Code has enterprise subscriptions. No $200 max plan but you can buy $100 premium seats.

(and if you are using per-token billing as an enterprise user without first maxing out a premium seat you are very silly)

New comment by fastball in "Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber"

fastball — Wed, 22 Jul 2026 06:44:33 +0000

I know many, many engineers who are paying something like 2-5% (via subscription) of what their usage would cost if billed by API tokens.

I know some down to about 1% ($200 Max plan vs $20k in tokens per month)

New comment by fastball in ""Drawing" the Mona Lisa with GPT-5.6, Claude, Gemini, and Grok"

fastball — Wed, 22 Jul 2026 06:33:20 +0000

The most interesting result for me is that they apparently prompted the models to optimize for SSIM, but many of the models trend worse over time. I suppose because viewing the canvas always comes after drawing, and they didn't give "revert to previous" capability as part of the toolkit.

Which in turn kinda jives with my experience of using these models for code: to some extent they only seem to have a concept of "forward", which invariably leads to "write more code to fix previous problems created", rather than taking a step back and removing broken things entirely.

New comment by fastball in "OpenAI and Hugging Face address security incident during model evaluation"

fastball — Wed, 22 Jul 2026 05:11:56 +0000

I'm sure that will change sooner rather than later, otherwise enterprising hackers will be able to claim that the model they were using went rogue.

New comment by fastball in "FreeInk: Open ecosystem for e-readers"

fastball — Tue, 21 Jul 2026 22:31:32 +0000

Oh yeah, indeed. Which is the rebrand haha

New comment by fastball in "FreeInk: Open ecosystem for e-readers"

fastball — Tue, 21 Jul 2026 19:10:17 +0000

Did you try FreeInk? I was debating which (between CrossPoint and this) to flash my new X3 with.

New comment by fastball in "You only need the frontier model for one single edit"

fastball — Mon, 20 Jul 2026 23:51:58 +0000

I don't think that is the takeaway at all.