Hacker News: drob518

New comment by drob518 in "One month coding with GLM 5.3 Flash"

drob518 — Sat, 03 Oct 2026 13:57:25 +0000

I sometimes switch mid-session. Depends what I’m doing. Sometimes a model is slow or it’s not giving me what I want and I don’t want to dump the context. But if there’s a natural break, I’ll start a new session with a new model.

New comment by drob518 in "One month coding with GLM 5.3 Flash"

drob518 — Sat, 03 Oct 2026 13:55:11 +0000

Those are my two workhorses as well.

New comment by drob518 in "One month coding with GLM 5.3 Flash"

drob518 — Sat, 03 Oct 2026 13:54:18 +0000

Cache cost is the primary metric, imo. Most people look at output cost, but you only pay that once. Cache cost you pay every turn and it grows each turn as the context length grows.

New comment by drob518 in "macOS Golden Gate Is a Buggy Mess"

drob518 — Tue, 29 Sep 2026 15:44:32 +0000

Well, even ignoring Claude’s capabilities, you would expect human testers (whether Apple employees or external beta testers) to be creating bug reports for all these things. It’s embarrassing if your self-proclaimed bug-fix/stability release is buggy and unstable. If the bugs are filed, then Claude should be able to help fix them.

New comment by drob518 in "Ember-1"

drob518 — Mon, 28 Sep 2026 18:29:56 +0000

Fine, let’s use your numbers, whatever they are. Either way, GLM 5.3 is a lot better than half as good as Kimi K3.

New comment by drob518 in "Ember-1"

drob518 — Sun, 27 Sep 2026 18:48:16 +0000

And GLM. Both Deepseek 4.1 Flash and GLM 5.3 Flash are quote verbose when thinking.

New comment by drob518 in "Ember-1"

drob518 — Sun, 27 Sep 2026 18:47:05 +0000

Unfortunately, it’s hard to make a chart of that.

New comment by drob518 in "Ember-1"

drob518 — Sun, 27 Sep 2026 18:45:53 +0000

Agreed. Even on the open weight side, GLM 5.3 has roughly equivalent performance to Kimi K3 for less than half the cost.

New comment by drob518 in "Reverse-engineering the Intel 8087's tangent algorithm: more than CORDIC"

drob518 — Sun, 27 Sep 2026 01:20:43 +0000

The combination of numerical methods and low-level hardware is fascinating. Those 20th century engineers were both smart and creative.

New comment by drob518 in "Disney+ and Hulu raise prices by up to 13 percent after doubling profits"

drob518 — Sun, 27 Sep 2026 01:03:48 +0000

You’re right. They’ll never be able to use their existing large revenue stream to fix it.

New comment by drob518 in "Disney+ and Hulu raise prices by up to 13 percent after doubling profits"

drob518 — Thu, 24 Sep 2026 16:22:27 +0000

All the major streaming providers are also adding live content. Once they have the infrastructure built out and the customers, it’s relatively easy to add the live content.

New comment by drob518 in "Disney+ and Hulu raise prices by up to 13 percent after doubling profits"

drob518 — Thu, 24 Sep 2026 16:19:38 +0000

Spot on. They discounted to encourage the transition away from cable and DVR (remember TiVo?) to streaming and then they raise prices to the magical “fraction of monthly budget for entertainment” level that everyone (cable, too) works up to. Every company is going to try to get you to give them all of what you’re willing to pay. When cancellations are greater than new subs, they’ll stop and wait for inflation and rising wages to catch up again. That’s the game. That’s always going to be the game.

New comment by drob518 in "Contrastive Language Models"

drob518 — Thu, 24 Sep 2026 16:09:17 +0000

Right, and I’d add that in some games there is actually a “win” state that serves as a binary indicator. Given this, you can actually view the model as the thing being classified into two states: (1) consistently wins the game and (2) doesn’t consistently win the game.

New comment by drob518 in "Owners mourn spoiled food after firmware update bricks Samsung smart fridges"

drob518 — Thu, 24 Sep 2026 14:22:22 +0000

Sigh. Anything + computer = computer

New comment by drob518 in "MiMo v2.6"

drob518 — Mon, 21 Sep 2026 22:18:22 +0000

Conspicuous that there’s no reference to GLM 5.3/Flash in the reported benchmarks. Just Deepseek and Kimi.

New comment by drob518 in "How OpenAI Used Its Own LLMs to Design Its Jalapeño Chip"

drob518 — Sat, 19 Sep 2026 14:32:49 +0000

Yea. That’s what happens when writers get sloppy.

New comment by drob518 in "How OpenAI Used Its Own LLMs to Design Its Jalapeño Chip"

drob518 — Sat, 19 Sep 2026 14:28:16 +0000

Hm. An interesting thought. Paired with RSI loops, that would allow rapid iteration in the hardware domain as well.

New comment by drob518 in "Fujitsu launches made-in-Japan next-generation CPU FUJITSU-MONAKA"

drob518 — Thu, 17 Sep 2026 14:32:01 +0000

One problem we’re going to have with AI hardware is coming up with a standard set of specifications that are comparable. I don’t really care about the CPU GHz and the memory bandwidth, at least not directly. What I really want to know is how many tokens per second this will deliver, but that also depends on the model. We need a standard metric for that. Perhaps we agree on a specific open weight model (e.g. GLM 5.3 Flash or Qwen vWhatever) and then measure TPS on the hardware of interest.

New comment by drob518 in "US confirms for first time it has deployed space weapons"

drob518 — Tue, 15 Sep 2026 19:10:23 +0000

When are we getting sharks with laser beams?

New comment by drob518 in "Will There Be a 7G?"

drob518 — Sat, 12 Sep 2026 19:56:19 +0000

Do we even need 6G?