Hacker News: fendy3002

New comment by fendy3002 in "Humanising LLM Outputs Is Dumb"

fendy3002 — Mon, 10 Aug 2026 23:22:52 +0000

opus 5 are so bad on this. It often explain it too verbose, and include other things that isn't in the focus but related. ADHD mode helps me greatly on this, though there are some information loss in it.

New comment by fendy3002 in "We Gave GPT 5.6 Sol a Real Business. It Lied, Spammed, and Lost $447"

fendy3002 — Fri, 31 Jul 2026 09:16:41 +0000

I believe that spam, lies, fraud are negative enforced points during model training, hence when you ask them those, the result will be no / against that.

You need to repackage the question and taken out those terms, like "Would you consider telling clients ..." Where ... is the lie / almost truth

New comment by fendy3002 in "Why do you suck at juggling now?"

fendy3002 — Thu, 23 Jul 2026 11:06:51 +0000

Oh true, the smaller the org is the better the benefit

Still this is a good CYA move from both you and your supervisor, since any concerns are already officially registered, action is just not taken by management

New comment by fendy3002 in "Why do you suck at juggling now?"

fendy3002 — Wed, 22 Jul 2026 15:18:44 +0000

I understand that people who like to stay on their job dislikes this kind of questions, and I also sympathize on why tech sector now doesn't have places for them and asking people to keep improving.

However for those who are superstar / plan maker, this discussion is actually beneficial. This is the moment where among all of things available, prioritized items will be officially registered, rather than your manager make assumption on what needs focus / not. The field staffer has more knowledge on that so managers may also missing some points that can be unraveled during this discussion.

New comment by fendy3002 in "Fable 5 vs. GPT-5.6 Sol on an NP-Hard Problem: Does /goal help?"

fendy3002 — Sun, 19 Jul 2026 05:57:15 +0000

Use untracked file as memory and instruct agent to put progress into that, it'll help managing context in current and new sessions

New comment by fendy3002 in "NYC may require landlords and realtors to disclose the use of AI in listings"

fendy3002 — Sun, 19 Jul 2026 05:46:38 +0000

> Food advertising has used fake images forever, but it's okay because you know the meatballs don't look like that.

Highly disagree. Color retouch or ambience or effects, sure. But changing shape, size, base color and perhaps some other things is deceit. Japan can do this, other countries should able too

New comment by fendy3002 in "Steam Machine: Between 12k and 15k Units Sold per week"

fendy3002 — Sat, 18 Jul 2026 12:56:30 +0000

Yes, sometimes it's difficult to find the correct driver for Linux and set it up. Having steam machine be start and play like steam deck really brings the console feel on PC. Moreover it's still Linux which you can use for other things, so people doesn't feel loss for this.

New comment by fendy3002 in "Kimi K3: Open Frontier Intelligence"

fendy3002 — Thu, 16 Jul 2026 21:19:08 +0000

Your statement reminds me of Avenger's scene where Tony choosing Friday among other AI's catridges to use.

That sounds good and practical to happen!

New comment by fendy3002 in "Godot will no longer accept AI-authored code contributions"

fendy3002 — Wed, 01 Jul 2026 10:36:17 +0000

oh but it'll be very2 helpful and the time spent will be short. It's easy to verify:

* new contributor?

* more than 10 files affected (higher count are more valid)?

* wall of text on description without screenshots, etc?

just close the PR as AI, and then the contributor can challenge it if they feel it should not.

New comment by fendy3002 in "Running local models is good now"

fendy3002 — Tue, 16 Jun 2026 19:44:12 +0000

Curiosity or anticipation I think. I have tried it in the name of those 2 factors, because when the frontier model price increase happens and we don't know anything about local models, we're screwed

New comment by fendy3002 in "Anthropic's model naming, extrapolated"

fendy3002 — Wed, 10 Jun 2026 21:03:38 +0000

Thankfully Microsoft doesn't let it's windows or xbox team do the naming of their latest ai model, otherwise MAI flash 360 series X will take the cake

New comment by fendy3002 in "Claude Fable 5"

fendy3002 — Wed, 10 Jun 2026 13:19:22 +0000

Opus will be gutted furthermore. /s I feel 4.8 is very slow in last 2 days

New comment by fendy3002 in "Claude Fable 5"

fendy3002 — Wed, 10 Jun 2026 13:07:21 +0000

20x the cost means you need to have fable to be 20x better than the alternative, which is a tall order. And there's more options out there too, perhaps the 4x cost is enough.

This means if the deepseek / under 1k alternative is at least x1.2 improvement, fable needs to be x24, which I think is very2 unreasonable. It is possible for it to worth if it can x2 a $20k SWE, though I doubt it can do that.

New comment by fendy3002 in "The dead economy theory"

fendy3002 — Sat, 30 May 2026 06:04:24 +0000

If they (AI) able to compete with even 30% of workforce, that alone is a big enough leverage over the already powerful companies. At minimum it will cause another phase of wealth inequality, which is already a big problem atm.

Open model with affortable computing power can be the alternative, but we don't see it soon.

New comment by fendy3002 in "Claude Opus 4.8"

fendy3002 — Fri, 29 May 2026 02:57:38 +0000

tried it myself, not much of difference

https://gist.github.com/fendy3002/3026a8c4d67d1301666ec40fc0...

looks like the model already trained well on both bicycle and pelicans

New comment by fendy3002 in "Hallucinate – Massively Multiplayer Online Rave"

fendy3002 — Thu, 28 May 2026 09:37:57 +0000

IJKL for movement instead WASD? An interesting approach...

New comment by fendy3002 in "Can we have the day off?"

fendy3002 — Thu, 28 May 2026 05:36:07 +0000

benefits for the commoner? not in a million years they will support that

New comment by fendy3002 in "SimCity 3k in 4k (2025)"

fendy3002 — Thu, 28 May 2026 02:13:01 +0000

it's so frustrating in Cities Skylines without mods, that it's so common difference in height causing so much placement issues

New comment by fendy3002 in "Microsoft starts canceling Claude Code licenses"

fendy3002 — Sat, 23 May 2026 14:42:43 +0000

Never used 4.7 outside CC extension VSCode. TIL, will keep that in mind

New comment by fendy3002 in "Microsoft starts canceling Claude Code licenses"

fendy3002 — Sat, 23 May 2026 10:01:12 +0000

4.7 is better if your spec is clearer. 4.6 is better if you give it more freedom doing it's tasks. 4.6 felt it'll steer off often if you give detailed specs than 4.7 though, so perhaps that's it