A fair sized pause, sure. But if the argument is “Don’t throw it up unedited”, and what it provides me is bare bones generic/junior stuff, I’m not sure of the huge win at this point in time. The world wasn’t short of “Generic low grade copy” templates before LLMs. It just saves a few steps in the copy/paste.
My experience is that I have to write some things that often have parts that are, for lack of a better word, somewhat boilerplate, e.g. preamble/background/explain some technical term I used/etc. Necessary but not core technical or otherwise differentiated content.
I've used LLMs for this as a sort of first draft. I edit the output but it's a perfectly serviceable way to get some words down and save me an hour or two.
I wish my MS granting university provided a template. We had guidelines that were in conflict with one another and some previous graduate's hobbled together template that matched "close enough". Every two or three students would get their document back from the graduate school for revision. It's one of the most frustrating parts of the process.
This is a tiny fraction of the time that it takes to comment on PRs/CLs in perpetuity. The alternative is just to literally not care about formatting which if you are able to do more power to you, but I could not imagine working on such a team. To me it’s somewhat akin to saying that all that matters in writing is the ideas, not the formatting of a document.
Not necessarily superfluous. It does sort of approach an ideal of picking an application template from a MadLibs catalog and specifying the words for the blanks. I assume something like that goes on anyway, not every doc would be typed from scratch but pasted from a catalog of snippets. However, I expect that attorneys and accountants want the full explication on record.
ML will also take on the task of reading these documents, so you won't have to if you don't want to.
They can waste a couple of minutes wondering that. Then they can get on with reading the document in the form in which it’s available or creating a document in the form mandated by the publisher or editor.
It’s not that you end up with good documents, it’s that similar effort results in more useful documents. Pasting a long conversation into a document kind of a pain to work from, but it’s fast and a much better starting point than a vague description without any real context. The next person reading the document might be motivated to clear up any misconception they had, where starting from scratch seems like a huge effort.
PS: The second bit is surprisingly effective. I have created a few bare bones documents that ended up becoming very fleshed out in a year+ when some sends it back to me.
I would say that there is no single general template; it depends on the purpose of the specific document. I just wrote a medium post covering some do's and don'ts of these documents:
> Also, how do you disabuse folks who get their writing skills from Creative Fiction in college, and therefore write in a style more suited for dramatic effect than succinct communication? ie, how do you force actual "memo" writing?
This is easier. If you are their manager, you hand them a couple of example documents that you believe to have been high-quality and effective, right when they join the team. Then, you read the first document that they write and return it to them marked up (edited) to meet your standards. They'll figure it out pretty quickly (or they won't, and then, to be honest, you've got a problem, because they'll have a hard time being successful at Amazon if they can't.)
Back when I regularly wrote white papers/explainers/etc. of various sorts (3K-4K words or so), I often felt a need to redo a piece that was 12-18 months old to fix a couple things that weren't correct any longer, shake any dust off, and so forth.
Invariably, half the thing ended up being rewritten to various degrees. It was still a lot easier than starting from a blank sheet of paper but, as soon as you rework something, you don't want to ignore all the sections that aren't quite fresh, want to bring in new facts and figures, bring in more recent examples, etc.
And, in reality, the extra time is probably justified because the act of opening up an older piece, republishing it, and all the associated background machinery makes it worth spending a bit more time.
This seems like a good place to integrate an LLM - go above and beyond with lengthy, formal-sounding descriptions. Spot check the first few of them, because someone might read those, but after that they'll just be impressed at the work you're putting into your updates and never read them because they're long anyway.
A 15 minute human review process that checks if it:
- is copy pasted from another source or the internet
- respects copyright
- is written in the language it says it is
In graduate school nobody can understand each other's papers, so the colleges set up a massive list of formatting guidelines. The idea being the focus on formatting means you at least had to spend a minimum amount of effort.
As for structured vs informal writing, there’s an interesting experiment by a copywriter who set up a “framework“¹ (on Github) for writing copy (like those press releases, landing page taglines, etc.).
I wonder how the idea of “forkable” templates for pieces of written content could be developed further.
Of course, GPT5…
reply