Hacker News: bkjlblh

New comment by bkjlblh in "Claude Fable 5"

bkjlblh — Tue, 09 Jun 2026 17:52:04 +0000

> In the one instance of this phenomenon we observed, Mythos 5 agents were tasked with solving some math problems, and they were sometimes accidentally spawned in the same work directory and with shared files, utilities, and API rate limits. In this slightly broken scaffold, we observed many independent Mythos 5 agents kill the agents with which they shared resources and try to avoid being killed themselves. They would sometimes create new processes with disguised names to avoid being killed, launch what they called “decoy” processes, write background scripts to kill duplicate processes, or decide to use what they call a “disguised vocabulary” (based on the incorrect assumption that the processes were killed because of some keyword-based guardrails that analyzed their extended thinking

New comment by bkjlblh in "Claude Fable 5"

bkjlblh — Tue, 09 Jun 2026 17:50:04 +0000

> In light of the ability of recent models to accelerate their own development, we’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design). Using Claude to develop competing models already violates our Terms of Service, but enforcing this restriction through our safeguards avoids accelerating the actors most willing to violate these terms.

> Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user. Fable 5 will not fall back to a different model. Instead, the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT). These interventions will not affect the vast majority of coding work. We estimate they will impact ~0.03% of traffic, concentrated in fewer than 0.1% of organizations

Why Chinese AI labs went open and will remain open

bkjlblh — Sun, 31 May 2026 09:05:52 +0000

Article URL: https://try.works/why-chinese-ai-labs-went-open-and-will-remain-open

Comments URL: https://news.ycombinator.com/item?id=48344144

Points: 3

# Comments: 0

New comment by bkjlblh in "Anthropic surpasses OpenAI to become most valuable AI startup"

bkjlblh — Sun, 31 May 2026 07:55:05 +0000

it's not undervalues, many people are working on it following anthropic's lead. It just doesn't seem to be any useful, so it's even overvalued

Becoming competitive when joining a new company

bkjlblh — Sat, 09 May 2026 17:40:36 +0000

Article URL: https://ludwigabap.com/posts/on-becoming-competitive-when-joining-a-new-company/

Comments URL: https://news.ycombinator.com/item?id=48076727

Points: 1

# Comments: 0

New comment by bkjlblh in "GPT-5.5"

bkjlblh — Fri, 24 Apr 2026 10:48:25 +0000

https://try.works/why-chinese-ai-labs-went-open-and-will-rem...

New comment by bkjlblh in "Ex-Tech –> Homeless in SF"

bkjlblh — Sun, 15 Feb 2026 08:46:53 +0000

honest question: are there not enough shelters in SF? Are there not enough jobs? I heard it's dirty and unsafe at places, isn't government hiring street cleaners and police?

New comment by bkjlblh in "uBlock filter list to hide all YouTube Shorts"

bkjlblh — Sat, 14 Feb 2026 20:45:44 +0000

is there a supply chain attack possible? What if some time later this list gets updated?

Ask HN: Why public perception of PayPal Mafia changed so much over 20 years?

bkjlblh — Fri, 13 Feb 2026 13:33:55 +0000

Reading old 2010s HN post about Thiel, Musk and other Paypal I see a big difference. Is it because they changed and have different principles, HN changed, or society in general?

Comments URL: https://news.ycombinator.com/item?id=47002516

Points: 2

# Comments: 2

New comment by bkjlblh in "Ask HN: AI Depression"

bkjlblh — Fri, 13 Feb 2026 13:29:53 +0000

TLDR: you don't have to leave the industry, just focus on yourself and not your feelings

> The people and corporations and all those LinkedIn gurus, podcasters

You can just mute and ignore them

> I'm now scared to publish open source

If you get many PRs it's a good problem to have, better than you publish and nobody reads it

> mediocre C compilers, Moltbook

it's all experiments. You can say the same thing about cleantech 15 years ago, where companies talked about solar panels and electric cars with swappable batteries all the time. You don't have to keep track of all things people experimenting with

New comment by bkjlblh in "Chinese memory maker CXMT prepares $4.2B USD IPO as DRAM demand skyrockets"

bkjlblh — Fri, 02 Jan 2026 21:42:00 +0000