The article links out to OpenAI's advice on prompting, but it also claims:

    OpenAI does publish advice on prompting o1, 
    but we find it incomplete, and in a sense you can
    view this article as a “Missing Manual” to lived
    experience using o1 and o1 pro in practice.

To that end, the article does seem to contradict some of the advice OpenAI gives. E.g., the article recommends stuffing the model with as much context as possible... while OpenAI's docs note to include only the most relevant information to prevent the model from overcomplicating its response.

I haven't used o1 enough to have my own opinion.

Those are contradictory. Openai claim that you don't need a manual, since O1 performs best with simple prompts. The author claims it performs better with more complex prompts, but provides no evidence.

The claims are not contradictory.

They are bimodal.

Bottom 20% of users can't prompt because they don't understand what they're looking for or couldn't describe it well if they did. This model handles them asking briefly, then breaks it down, seeks implications, and prompts itself. OpenAI's How to Prompt is for them.

Top 20% of users understand what they're looking for and how to frame and contextualize well. The article is for them.

The middle 60%, YMMV. (But in practice, they're probably closer to bottom 20 in not knowing how to get the most from LLMs, so the bottom 20 guide saves typing.)

I'm not saying it won't work. I'm just asking for evidence. You don't think its strange that none of the authors or promoters of this idea provided any evals? Not even a small sample of prompt/response pairs that demonstrate the benefit of this method?

In case you missed it

    OpenAI does publish advice on prompting o1, 
    but we find it incomplete, and in a sense you can
    view this article as a “Missing Manual” to lived
    experience using o1 and o1 pro in practice.

The last line is important

But extraordinary claims require extraordinary proof. Openai tested the model for months and concluded that simple prompts are best. The author claims that complex prompts are best, but cites no evidence.

Requiring only simple prompts surely sells better. I would not assume the documentation provided by OpenAI is totally unbiased and independent of business goals.

I find it surprising that you think documentation issues are “extraordinary”.

You have read literally any documentation before, right?

I'd just love to see one of the prompt/response pairs demonstrating the technique. Here is an example from my llm-consortium tool. If you want more, you can have them in 5 minutes. https://github.com/irthomasthomas/llm-consortium/blob/main/c...

I mean, OpenAI not only tested the model, they literally trained the model. Training a model involves developing evaluations for the model. It’s a gargantuan effort. I’m fairly certain that OpenAI is the authority on how to prompt o1.

I'm not sure what this comment is trying to insinuate? I mean sure they might have a financial interest, but they're also writing about stuff they know well since they use it.

Basically that talk about and use JavaFX are those with a vested interest in JavaFX as a platform in and of itself. Rather than hearing from people who are more interested in the programs that you'd make with JavaFX.

The connotation is that there isn't much adoption of JavaFX outside of a small (Sun/Oracle-centric) group, which has largely always been the case with JavaFX.

I'll add my two cents on top of the author's. I have both the Amazon Echo and the Amazon Tap. This article was spot on. I'm perhaps less forgiving than the author. If Alexa can't answer a question on the first try I'll just give up. When Alexa does answer a question or a request she usually does a phenomenal job and it's just really cool. It really feels futuristic.

Echo is cool, it can be pretty useful, but it hasn't realized its full potential yet.

I bought the Tap as a portable speaker to take with me on bike rides with friends. Well I took it out the other day and the minute it lost WiFi connection it just wouldn't work. I'm sure this was a problem on my end, but I had already paired the devices and I wasn't going to spend time fiddling with it while I was out with friends.

EDIT: Also it doesn't happen too often but Alexa has definitely accidentally triggered. It even woke me up at 2 AM after the power went out temporarily to tell me that it had no WiFi.

We've only had our Echo a couple weeks, but we've already found that characters named Alexa in TV shows are a problem, and that at least once it has responded to something that wasn't "Alexa" from the TV. So yeah, it definitely has false positives.

I got to take Matt's compilers course. He was a really exceptional professor who had a way of presenting pretty mundane and rigorous material in an interesting way. His lectures were always engaging. If your interested in understanding programming languages at a lower level I can't recommend enough perusing his blog.

He addresses this in his post.

  if x.in? my_set   # very ugly

With that said, I personally don't agree with the post.

I don't know what's so ugly about this.

I wouldn't call the method "in?" but maybe "element_of?" or "member_of?" but besides that, I don't see the need for an operator.

Thank you for this quote. I often think that this gets overlooked in our industry, and I'm no exception. A recent Hanselminutes podcast[1] touched on this. It pointed to a great resource[2].

1. http://hanselminutes.com/475/accessibility-a11y-with-opendir...

2. http://a11yproject.com/

I'm a current CS student, finishing this year. I work with both people who are self taught and people who have degrees. I think pretty highly of those I work with who are self taught. All in all the differences are minimal when getting the job done.

CS emphasizes theory and mathematics. Programming and coding is usually discussed in the introductory courses, but from there students are expected to teach themselves to keep up.

This was a really cool read :). I'm working on my own DSL for generating code. I'll make a Show HN soon, but it was really cool to see someone else doing something so similar.

Nice what kind of code is it going to be used for to generate? A specific language or something more generic.

It would be cool with a DSL to generate low level code like X86 ASM, LLVM IR or JVM bytecode. That would be so meta!

Heh, not so complicated. It's just used to generate boiler plate code. Kind of like data object files in C#, Java, PHP, etc. That would be really cool though :).

I'm taking the class this upcoming semester from Matt. I'm excited to understand what tools like gcc, clang, and so on are doing under the hood.

I'm looking forward to having you in the class!

I agree with the idea of motivation. An anecdote for the OP:

In the past when trying to workout it would generally go something like this:

* Immediate motivation, go strong for 1-3 weeks.

* Work or lack of interest interferes with workout regimen.

* Eventually interest dies out and I stop going.

What helped me break this cycle was finding motivation by:

* Doing things I liked to supplement the 'work' part. This meant playing racquetball, ultimate frisbee, and cycling while lifting.

* Learning about the benefits beyond looking physically attractive.

* Going consistently that a habit has formed.

Seriously, I feel different if I miss my gym routine for a day. Hope it helps as you think about your own material to write about.