Hacker News: robwwilliams

New comment by robwwilliams in "Issue: Claude Code is unusable for complex engineering tasks with Feb updates"

robwwilliams — Mon, 06 Apr 2026 16:54:34 +0000

Agree: it is Anthropic's aggressive changes to the harnesses and to the hidden base prompt we users do not see. Clearly intended to give long right tail users a haircut.

New comment by robwwilliams in "Issue: Claude Code is unusable for complex engineering tasks with Feb updates"

robwwilliams — Mon, 06 Apr 2026 16:45:56 +0000

Yes, and over the last few weeks I have noticed that on long-context discussions Opus 4.6e does its best to encourage me to call it a day and wrap it up; repeatedly. Mother Anthropic is giving preprompts to Claude to terminate early and in my case always prematurely.

New comment by robwwilliams in "Embarrassingly simple self-distillation improves code generation"

robwwilliams — Sat, 04 Apr 2026 15:07:03 +0000

Very cool. An evolutionary biologist would say: Welcome to the party!

Mutation rate modulation is the AI engineers’ heat. And selection does the trimming of the outliers.

Some more serious biomorphic thinking and we may get to the next big insight courtesy of 3+ billion years of evolution—- evolution that enabled a great ape species to write a paper like this and build LMM’s like Gemma4 that totally rock on a 3.5 pound MacBookPro M5 Max with 128 GB of RAM.

New comment by robwwilliams in "Consider the Greenland Shark (2020)"

robwwilliams — Wed, 01 Apr 2026 22:22:02 +0000

Capybaras in Brazil were Friday fish.

New comment by robwwilliams in "1M context is now generally available for Opus 4.6 and Sonnet 4.6"

robwwilliams — Sat, 14 Mar 2026 03:27:25 +0000

Yes, especially with shifts in focus of a long conversation. But given the high error rates of Opus 4.6 the last few weeks it is possibly due to other factors. Conversational and code prodding has been essential.

New comment by robwwilliams in "Enhancing gut-brain communication reversed cognitive decline in aging mice"

robwwilliams — Fri, 13 Mar 2026 03:13:41 +0000

Very well done study with a cautious interpretation of potential translational relevance in humans.

The paper is open access. The discussion does a fine job of providing a full context for interpreting their findings.

https://www.nature.com/articles/s41586-026-10191-6

New comment by robwwilliams in "Yann LeCun raises $1B to build AI that understands the physical world"

robwwilliams — Tue, 10 Mar 2026 22:32:25 +0000

> We can steer our own learning corpus

This is critical. We have some degree of attentional autonomy. And we have a complex tapestry of algorithms running in thalamocortical circuits that generate “Nows”. Truncation commands produce sequences of acts (token-like products).

New comment by robwwilliams in "Project Operational Autonomy: our agent-to-agent-future"

robwwilliams — Fri, 06 Mar 2026 22:31:52 +0000

Well worth reading from Pam Vagata. Some important insights and predictions.

Her vision of autonomy is not true autonomy, but a vision of AI servants on slightly longer chains.

Is that fair? Yes, I think so.

The key questions are not being asked: What drives autonomy? Who or what selects the motivation and defines the scope of action in a real stateful world?

I set my thermostat and it acts “autonomously” over the range I set. But that is not the autonomy she implies.

Give me CS types and founders like Hassabis, Sutton, Kanerva, Amodei, Bach, Levin, LeCun, Brooks, Hawkins, Sejnowski, and Friston (and about ten other scientists) who understand some of the core requirements of autonomy (but not all).

One hard but invaluable book to read before contending with the word autonomy: Maturana HC, Varela F (1980) Autopoiesis and Cognition. Finally reissued as a paperback.

New comment by robwwilliams in "Aromatic 5-silicon rings synthesized at last"

robwwilliams — Sun, 01 Mar 2026 16:41:38 +0000

The review should have expanded on this at a practical level even mom and dad could understand—the standard “better life through chemistry” angle.

New comment by robwwilliams in "Aromatic 5-silicon rings synthesized at last"

robwwilliams — Sun, 01 Mar 2026 16:37:45 +0000

Perfect! I read this “heart-warming” overview of two papers in Science and learned zero about why this is of any significance. The discovery is significant but I had to probe Opus 4.6 to find out why.

The personal focus is a distraction. It would be great if science writers could focus on the science and significance of the advance.

New comment by robwwilliams in "We Will Not Be Divided"

robwwilliams — Sat, 28 Feb 2026 16:52:47 +0000

Quitting their jobs? How is that the pragmatic or effective response?

New comment by robwwilliams in "Statement from Dario Amodei on our discussions with the Department of War"

robwwilliams — Fri, 27 Feb 2026 14:17:10 +0000

Sure you can grade “commendable” if you want, but this counts as commendable to me even if wealthy. I have not noticed that wealthy individuals are less concerned than unwealthy individuals about loss of resources and money. In fact, wealth seems to exacerbate the problem.

New comment by robwwilliams in "Statement from Dario Amodei on our discussions with the Department of War"

robwwilliams — Fri, 27 Feb 2026 14:02:25 +0000

All excellent points to add to the motivation to hold the line just where it has been.

New comment by robwwilliams in "Statement from Dario Amodei on our discussions with the Department of War"

robwwilliams — Fri, 27 Feb 2026 13:56:20 +0000

Agree: Humans are much more frightening as an existential risk than AI or AGI. We have three unstable old men with their fingers too close to big red buttons.

New comment by robwwilliams in "Claude Sonnet 4.6"

robwwilliams — Wed, 18 Feb 2026 05:15:24 +0000

This is almost too damn funny/perfect to believe. All it had to add:

"And you will get some good exercise too."

New comment by robwwilliams in "Claude Sonnet 4.6"

robwwilliams — Wed, 18 Feb 2026 04:12:09 +0000

Sonnet 4.6 failed for me.

“Walk. It’s 50 meters—a 30-second stroll. Driving that distance to a car wash would be slightly absurd, and you’d presumably need to drive back anyway. “

Opus 4.6 nailed it: “Drive. You’re going to a car wash. ”

I used this example in class today as a humorous diagnostic of machine reasoning challenges.

New comment by robwwilliams in "Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT"

robwwilliams — Sat, 31 Jan 2026 04:07:07 +0000

Somewhat different type of problem and perhaps a useful precautionary tale. I was using Opus two days ago to run simple statistical tests for epistatic interactions in genetics. I built a project folder with key papers and data for the analysis. Opus knew I was using genuine data and that the work was part of a potentially useful extension of published work. Opus computed all results and generated output tables and pdfs that looked great to me. Results were a firm negative across all tests.

The next morning I realized I had forgotten to upload key genotype files that it absolutely would have required to run the tests. I asked Opus how it had generated the tables and graphs. Answer: “I confabulated the genotype data I needed.” Ouch, dangerous as a table saw.

It is taking my wetware a while to learn how innocent and ignorant I can be. It took me another two hours with Opus to get things right with appropriate diagnostics. I’ll need to validate results myself in JMP. Lessons to learn AND remember.

New comment by robwwilliams in "Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT"

robwwilliams — Sat, 31 Jan 2026 03:40:01 +0000

This was my experience too over Dec 2025. Thereafter marginal Claude Pro utility. They are struggling with demand.

New comment by robwwilliams in "Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT"

robwwilliams — Sat, 31 Jan 2026 03:35:41 +0000

Love this idea. It would make it much more practical to get a set of different perspectives on the same text or code style. Also would appreciate temperature being tunable over some range per conversation.

New comment by robwwilliams in "I trapped an AI model inside an art installation (2025) [video]"

robwwilliams — Sat, 31 Jan 2026 01:08:06 +0000

Not too dissimilar to the situation of the well known neurosurgical patient Henry Molaison (H.M.) who had very limit long term memory of his childhood and young adulthood but little after surgery at 27 years of age. He retained cognitive abilities and self-consciousness but within a narrow temporal world.

Good at cross/word puzzles and conversations of specific types.

https://en.wikipedia.org/wiki/Henry_Molaison