Hacker News: port3000

New comment by port3000 in "Claude Opus 5.5"

port3000 — Tue, 22 Sep 2026 16:50:34 +0000

I believe Opus 5 isn't meant to be spoken to by humans. It's great at executing but I reckon it's intended to be spoken to by other models such as Fable. I use Fable as the orchestrator, only speak with Fable, and all implementation, recon, design etc happens with Opus 5, with Fable reviewing (and translating).

New comment by port3000 in "Can we stop with the uptime percentages?"

port3000 — Wed, 16 Sep 2026 16:18:13 +0000

Until there is an industry wide definition of outage, degraded performance, etc then it's all moot

New comment by port3000 in "Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling"

port3000 — Mon, 20 Jul 2026 16:13:19 +0000

It's amazing how quickly Fable went from 'Game-changing model that needs to be banned' to 'Yeah it's alright, but OpenAI is also just as good and there are a couple of good open weight alternatives that are equivalent for almost everything'

The hype cycles are shortening, perhaps we really are reaching some kind of plateau this time (famous last words)

New comment by port3000 in "Codex is wearing out our devices"

port3000 — Mon, 20 Jul 2026 14:34:15 +0000

Interesting, thanks, had no idea

New comment by port3000 in "Codex is wearing out our devices"

port3000 — Mon, 20 Jul 2026 14:28:54 +0000

Does writing to an SSD often really 'wear out' the device?

New comment by port3000 in "Kimi K3, and what we can still learn from the pelican benchmark"

port3000 — Fri, 17 Jul 2026 18:51:58 +0000

Yeah I asked Nano Banana to make a render of our company office and was scarily accurate

New comment by port3000 in "AWS: Inaccurate Estimated Billing Data – $1.7 billion"

port3000 — Fri, 17 Jul 2026 11:41:03 +0000

Rookie error

New comment by port3000 in "AWS: Inaccurate Estimated Billing Data – $1.7 billion"

port3000 — Fri, 17 Jul 2026 11:40:02 +0000

They have to pay for that AI Capex buildout somehow

New comment by port3000 in "Chinese AI startup Moonshot to launch model challenging Anthropic's lead"

port3000 — Thu, 16 Jul 2026 12:00:33 +0000

Paywall, but exclusive from the FT:

Kimi K3 launching in the coming days

Expected to exceed Opus 4.8 in capability

Moonshot raising new funds valuing company above 30bn

New comment by port3000 in "Codex Micro"

port3000 — Wed, 15 Jul 2026 16:41:57 +0000

Where have I seen this before

https://www.reddit.com/r/ClaudeCode/comments/1ue5inx/i_built...

New comment by port3000 in "I tricked Claude into leaking your deepest, darkest secrets"

port3000 — Wed, 15 Jul 2026 15:19:57 +0000

Oh...well I am a Silly Bean!

New comment by port3000 in "Apple's new SpeechAnalyzer API, benchmarked against Whisper and its predecessor"

port3000 — Wed, 15 Jul 2026 08:13:30 +0000

If you use the offline-only models (I use the Nvidia one and it's great) and put it in offline mode it is. I've used it for 6 months and haven't paid anything

New comment by port3000 in "I tricked Claude into leaking your deepest, darkest secrets"

port3000 — Wed, 15 Jul 2026 08:01:01 +0000

My name in Claude is Silly Bean. I did it at first because it made me chuckle every time I opened Claude and it said 'Back again, Silly Bean?'

But turns out I was playing 4D cybersecurity chess

New comment by port3000 in "Apple's new SpeechAnalyzer API, benchmarked against Whisper and its predecessor"

port3000 — Mon, 13 Jul 2026 23:11:13 +0000

Thanks will check it out, hadn't heard of it

New comment by port3000 in "Apple's new SpeechAnalyzer API, benchmarked against Whisper and its predecessor"

port3000 — Mon, 13 Jul 2026 17:02:24 +0000

I use Spokenly, offline-only mode with the Nvidia model. All local, totally free. Highly recommend

New comment by port3000 in "GLM 5.2 and the coming AI margin collapse"

port3000 — Tue, 07 Jul 2026 09:34:25 +0000

Interesting, so point 2 means that a lot of the hardware being installed now won't be able to run the frontier models of 2029? How does that change the demand for compute/models in the future, I can imagine that even if OpenAI/Anthropic have a moat in 2029 there will be so much older hardware and such a hangover from that investment boom that there will be very little installed capacity that can run it

New comment by port3000 in "Claude Code is steganographically marking requests"

port3000 — Tue, 30 Jun 2026 16:57:44 +0000

That's a lot of effort when they could just play a short video saying 'You wouldn't steal a car' instead

New comment by port3000 in "GLM 5.2 beats Claude in our benchmarks"

port3000 — Mon, 29 Jun 2026 07:20:36 +0000

Opencode Go subscription ($5 to try for one month) or Neuralwatt are what I use. Both through opensource Opencode harness (like Claude code)

New comment by port3000 in "Claude Fable 5: mid-tier results on coding tasks"

port3000 — Fri, 12 Jun 2026 08:54:36 +0000

It's just much more thorough and spins up a lot of subagents to basically do a lot more E2E testing. Not necessarily smarter, imo you could get the same result with a lesser model by procedurally prompting, but a lot more compute and orchestration.

New comment by port3000 in "Claude Fable 5: mid-tier results on coding tasks"

port3000 — Fri, 12 Jun 2026 08:50:16 +0000

Fable feels like a slightly more advanced 4.5/4.6 (less verbose than 4.7 and 4.8) with more adversial work checking. And a lot more compute to be more thorough from the first prompt. I feel it would be possible to get pretty much the same results with 4.6 with enough back and forth iterations. It kind of makes sense to me that this is the 'magic' behind Mythos and its cyber capabilities too. Just a massive iterative loop and really going into a lot more detail on edge cases.