Hacker News: smcleod

New comment by smcleod in "Changes at Google DeepMind: Demis Hassabis from CEO to Chair, Jeff Dean departs"

smcleod — Wed, 05 Aug 2026 23:04:20 +0000

I think it's a bit inflated to label Gmail a "great tech for consumers". Maps certainly, YouTube mostly, Chrome & Android for loyalists, but Gmail hasn't been "great tech" for well over a decade.

Additionally these are all old products - even the company's more recent products such as Gemini feel stale and on unsteady ground.

New comment by smcleod in "LLMs reward expertise"

smcleod — Tue, 04 Aug 2026 13:32:53 +0000

Similarly there is research that shows the quality of LLM outputs strongly correlate with the education level (in the field) of the person promoting them.

New comment by smcleod in "KOReader"

smcleod — Wed, 29 Jul 2026 13:28:22 +0000

Japanese software and software technology culture is amongst the worst in the world. They're famous for failing to adopt modern ways of working, tools and as an industry is seriously struggling. There's fantastic video in this from Asianometry: https://youtu.be/ky1nGQhHTso

New comment by smcleod in "LG to ban residential proxies from smart TV apps"

smcleod — Wed, 22 Jul 2026 12:33:26 +0000

You know your phone is OLED right? I'm assuming you left a still image on without the inbuilt screen refresher enabled or something? I remember seeing that way back in the early OLED days but not on good TVs in the last 8~ years.

New comment by smcleod in "LG to ban residential proxies from smart TV apps"

smcleod — Wed, 22 Jul 2026 09:29:33 +0000

Really? I've had many updates over the 7 years since my c9 was released, still getting them as recent as earlier this year I think.

New comment by smcleod in "LG to ban residential proxies from smart TV apps"

smcleod — Wed, 22 Jul 2026 07:11:42 +0000

7 years into my C9 and it never fails to impress.

New comment by smcleod in "LG to ban residential proxies from smart TV apps"

smcleod — Wed, 22 Jul 2026 07:10:50 +0000

If you care about home cinema it's important. There's a pretty big jump down. You can look up the specs and reviews of the C and G series from LG. Anecdotally quite a few times when people see my 7 year old LG C9 I've had them ask if it was some new expensive TV as they are so impressed with the picture.

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

smcleod — Wed, 22 Jul 2026 04:57:46 +0000

I suspect the entire 3.x family is fundamentally problematic and we'll need to see an architecture change before they're half decent like back in the 2.5 days again.

New comment by smcleod in "LG to ban residential proxies from smart TV apps"

smcleod — Wed, 22 Jul 2026 04:56:12 +0000

I'm aware of that setting, but this is the Samsung baseline. The panels they export for OEMs / other manufacturers don't seem to have the same hyper saturation. Even just standing next to someone using a Samsung phone or tablet you can often spot it straight away we know you what to look for.

New comment by smcleod in "LG to ban residential proxies from smart TV apps"

smcleod — Wed, 22 Jul 2026 04:09:48 +0000

Their panels are often over-saturated with that fake HDR like look, similar to their phones.

New comment by smcleod in "LG to ban residential proxies from smart TV apps"

smcleod — Wed, 22 Jul 2026 03:50:04 +0000

Unfortunately they make arguably the best OLEDs and webOS is pretty decent to use (especially compared to the terrible OS that Samsung and Sony run). I think the best thing people can do is update their firmware then disconnect it from the internet. Using an AppleTV or similar provides a better experience than any TVs built in apps anyway.

New comment by smcleod in "OpenAI reduces Codex Model Context Size from 372k to 272k"

smcleod — Sun, 19 Jul 2026 19:46:54 +0000

I haven't used Codex in some time, but in Claude Code I force my context size down to 280k on purpose as in my experience your best work is done within the 30-250k~ range regardless of a models maximum context size.

New comment by smcleod in "Kimi K3: Open Frontier Intelligence"

smcleod — Thu, 16 Jul 2026 21:10:00 +0000

It's 2.8T, I'm sure they will open the weights but it will only be able to be run on very high end machines.

New comment by smcleod in "Kimi K3: Open Frontier Intelligence"

smcleod — Thu, 16 Jul 2026 21:03:26 +0000

And American providers, not sure if it's still the case but OpenAI were doing this.

New comment by smcleod in "The lost joy of music piracy"

smcleod — Thu, 16 Jul 2026 06:53:37 +0000

Soulseek is still going hard.

New comment by smcleod in "The classifiers Anthropic puts in front of Fable are too zealous"

smcleod — Thu, 09 Jul 2026 00:38:55 +0000

I suspect there's some cargo culting, and some folks that are generally more likely to table flip than understand things that challenge their workflow.

New comment by smcleod in "The classifiers Anthropic puts in front of Fable are too zealous"

smcleod — Thu, 09 Jul 2026 00:37:04 +0000

Yes their docs cover how to properly prompt it to ensure you don't get hit by the classifier, you can distil those down to a skill to help build prompts for the more challenging situations, e.g. https://github.com/sammcj/agentic-coding/blob/main/Skills/pr...

New comment by smcleod in "The classifiers Anthropic puts in front of Fable are too zealous"

smcleod — Wed, 08 Jul 2026 21:22:03 +0000

I've been working with it heavily since its first release. I use it for software architects, complex debugging and some development and I have not had it refuse or downgrade even once.

New comment by smcleod in "AMD Ryzen AI Halo – $4k AI Dev Kit"

smcleod — Mon, 06 Jul 2026 21:24:36 +0000

Yeah the 256GB/s bandwidth is really very limiting

New comment by smcleod in "GPT-5.6 Sol Ultra will be in Codex"

smcleod — Mon, 06 Jul 2026 09:45:36 +0000

ultracode in Claude Code kicks off a dynamic workflow.