clbrmbr

6 ms·

61.
by clbrmbr 5mo ago
I’ve had great success structuring requirements using Johnny Decimal. Well organized and numbered requirements for each project is mega helpful for agents and humans throughout the progress. Oh and I love açaí.
62.
by clbrmbr 6mo ago
They also (at the cantonal level) have disparate education systems, with classes and grade levels mismatching between neighboring cantons. Yet, if you check what typical Swiss high school students are actually leaning (say at College de Can
63.
by clbrmbr 6mo ago
Switzerland is apparently a federation of federations. Local self-determination. Amazing place if you ask me would move back there in a heartbeat.
64.
by clbrmbr 6mo ago
How can they change the tokenizer without a wholesale pre-train?
65.
by clbrmbr 6mo ago
And all those Delphi programs (ok rn I can only think of the crackz but there must have been others). What made these Delphi programs so unique in their UIs?
66.
by clbrmbr 6mo ago
There’s also a paper [0] from many well known researchers that serves as a kind of informal agreement not to make the CoT unmonitorable via RL or neuralese. I also don’t think Anthropic researchers would break this “contract”. [0] https:&#
67.
by clbrmbr 6mo ago
But still CoT distillation WORKS. See the DeepSeek R1 paper.
68.
by clbrmbr 6mo ago
> There is no cohort of senior product leaders who developed their judgment in conditions where their teams were expected to demonstrate financial return, because those conditions did not exist during the years when that cohort was learn
69.
by clbrmbr 6mo ago
25A removal is temporary pending a bar in congress even higher than that for impeachment (2/3 of house and senate).
70.
by clbrmbr 6mo ago
> OpenAI was on the verge of closing a large investment from Thrive, a venture-capital firm founded by Josh Kushner, Jared Kushner’s brother, whom Altman had known for years. The deal would value OpenAI at eighty-six billion dollars and
71.
by clbrmbr 6mo ago
A few tricks to beat this are community, brand, an app that only works with the official version (hard to pull off and sad), and going more niche.
72.
by clbrmbr 7mo ago
@dang Can a moderator update the link? The original is much better and we shouldn’t promote the copyposter.
73.
by clbrmbr 7mo ago
I once was a communications contractor for the major NJ power utility. One of their long time field techs (let’s just refer to him as Mr. T) was giving a tour of a substation that was built from the looks of it in the 50s. I have, you see,
74.
by clbrmbr 7mo ago
Page not rendering well on iPhone Safari. Good content tho!
75.
by clbrmbr 7mo ago
I strongly object to building weapons. It is not right. Raise your consciousness, young hacker. I grew up building homemade rocket engines to power model rockets. I even programmed a flight computer in ASM. I was always quite risk averse an
76.
by clbrmbr 7mo ago
Do you have a more detailed trace that shows the system reminders? Do you know in which order it was fed into the LLM call that resulted in the bad reasoning?
77.
by clbrmbr 7mo ago
Huh. I’m missing out I guess. Is there a plugin you use for spinning them up? Heavy superpowers/CC user here.
78.
by clbrmbr 7mo ago
Not everyone. If your business is chill and you are REEEEALY thoughtful and respectful with newsletters you will be rewarded with open rates well in excess of 50%…
79.
by clbrmbr 7mo ago
Hahah yeah if you play with LoRas on local models you will see this a lot. Most often I see it hallucinate a user turn or a system message.
80.
by clbrmbr 7mo ago
This. The models struggle with differentiating tool responses from user messages. The trouble is these are language models with only a veneer of RL that gives them awareness of the user turn. They have very little pretraining on this idea o
81.
by clbrmbr 7mo ago
Same here. Auto mode is NOT ok. Sadly, smaller models cannot be trusted with access to Bash.
82.
by clbrmbr 7mo ago
man i miss GPT-4.5
83.
by clbrmbr 7mo ago
Thanks. Do come back to post with your tutorials. I'd recommend going quite granular and being didactic. Take somebody who understands gradient descent on MLPs and ELI5 the analytical part. Try to anticipate some of the doubts (there w
84.
by clbrmbr 7mo ago
Here to say I'm one of those people who did my first Show HN recently, and it was 100% due to the lowered activation energy to build something awesome with Claude. Not 45min, but took about 6 hours of my time, and benefitted from test
85.
by clbrmbr 7mo ago
I’d love to see a post that clearly walks through how this works for some examples to give the intuition. And then, how much is really saved on training for a non-trivial model? And is this applicable to deep models?
86.
Show HN: Treemap Firmware Bloat Visualizer (Rust/WebASM)(merck.substack.com)
1 points by clbrmbr 7mo ago|0 comments
87.
by clbrmbr 7mo ago
This would be awesome. Even titles and shasums could be enough.
88.
by clbrmbr 7mo ago
Just deleted my account. Sama’s bad faith communication on X really sealed the deal for me. Moving my company over to Claude on Monday.
89.
by clbrmbr 8mo ago
I suggest reading the tests that Superpowers author has come up with for testing the skills. See the GitHub repo.
90.
by clbrmbr 8mo ago
AmaZing job. My 6 and 8 yo could understand back to 1400.

More ›