_cjk

|

[_cjk @_colinjk]

@majamediaco There’s something in generative AI being the opposite of generative, because of the lack of constraint

Like an inverse theory of constraints, where rather than seeking to remove them, you seek to choose them

[_cjk @_colinjk]

Micro-skill to pan chats for gold

It's a minimal loop: search for what matters, tell you, run another pass

https://t.co/n6825jzTGb

Claude Code / Codex uses subagents to orchestrate multiple passes into one result.

[_cjk @_colinjk]

@vixsheikh I like this line of thinking but what would instigate change? What’d make someone decide to hire a coach? Other people doing it?

Maybe thinking jobs is one step ahead of what’s real

What problems are undersolved? What value is not realised? What problems and value are unknown

[_cjk @_colinjk]

“Now, here, you see, it takes all the running you can do, to keep in the same place.”

[Thorsten Ball @thorstenball]

"If you stop, you're dead"

Some thoughts on model providers, the race they're in, inference margins, and the tectonic forces shaping software right now.

[_cjk @_colinjk]

@thorstenball You’ll walk in a field talking to yourself and you’ll be happy. (Love this do more.)

[_cjk @_colinjk]

@CritThought94 @CinemaTweets1 @paramountplus GOL S1 is something else, wish Gareth Evans stayed involved. Yeah, didn't finish S3. Not gangster but The Jackal was good. The Agency is good. Again not crime but similarity is operators.

[_cjk @_colinjk]

@CritThought94 @CinemaTweets1 @paramountplus Hahaha I took for you a connoisseur, fair enuf

[_cjk @_colinjk]

@CritThought94 @CinemaTweets1 @paramountplus Yeah. What hasn’t disappointed in that vein?

[_cjk @_colinjk]

@CritThought94 @CinemaTweets1 @paramountplus Thought the presence of mind, three steps ahead, relentless 24/7 almost made Harry a separate thing to the whole. Like I could happily watch a Harry edit. What I meant was: what’s better? Fishing for tips

[_cjk @_colinjk]

@CritThought94 @CinemaTweets1 @paramountplus Gangster fantasy but loved it personally. What is your touch point for good?

[Behzod @beh_zod]

This is because there are three types of writing, based on the primary relationship between two of author, content, and audience, and AI only feels good for one of them right now.

Technical writing is about the relationship between content and the audience. In some cases there is an objectively correct set of things to write (like a dictionary definition), but in many other cases, the writing needs to be correct more than anything else (enjoyable, stylistic). Code falls into this category (putting aside the arguments about style in coding for a moment). You want code that works, so you're okay with AI writing code because it's about the creation of correct answers.

Communication is about relationship between the author and audience, and using AI here often feels like a betrayal because you're mediating that relationship in ways that you (as the audience) did not invite. Furthermore, because there is not a technically correct answer, there is only a shared journey of understanding, using AI without your permission feels like a one-sided journey. This is where you are having the most visceral (and deserved) reaction.

Now, if you believed that someone was using AI to help aid in the journey (not shortcut it), that's a different story. But it feels like (at least for me) that people use AI when they want the outcome without the work, and the entire point of communication is the work of developing shared understanding. AI that serves the work = great. AI that shortcuts the work = terrible.

(Art is about the relationship between author and content, and I think we can skip that for now.)

[_cjk @_colinjk]

@sushrutabasak it’s a risk to express yourself. most people don’t. i don’t know the ratio of contribitor vs reader on x, but imagine it is 1000:1. imagine a lot of ai writing starts with someone asking the ai what it thinks
about a draft, and it seeming safer, “better” or disassociated risk

[_cjk @_colinjk]

@thorstenball Scared you're going laugh at this but fuck it.

Make like it's '23 with this:

https://t.co/H8cS9EPTXN

[_cjk @_colinjk]

@theslowtell @Afinetheorem @logangraham Points to a near future of tasteful, synthetic but personalised, RLHF? LinkedIn taste echo chamber as the right diagnosis, but wrong direction

[_cjk @_colinjk]

@majamediaco @lulumeservey "We must resist becoming intellectually sedentary" is the magic spell for me

is fascinating

both how AI impacts us (and similar, how TV and reading and other mediums did and do); and how to intentionally resist, while also leaning in

[_cjk @_colinjk]

The first rule of Good Taste Club is you do not talk about good taste

[_cjk @_colinjk]

@rcolvile @v_j_freeman I’ve not heard the positive case for water privatisation before. I get that’s my bubble. Many in the public hold Burnham’s views. Others will find them plausible. Anti privatisation meme is strong, ancient gospel. How can this change?

[_cjk @_colinjk]

In the land of the slop, the half-assed man is king

[_cjk @_colinjk]

Opus has a real tendency to play shell games with the point

"Your message named 1 and 3; the piece to keep distinct is 2"

"It's three things wearing the coat of fourth"

Head hurts just trying to read the floral slop

[_cjk @_colinjk]

@mitsuhiko Anthropic are clearly an American company. EU/UK need their own Anthropic. The parallel is France and nuclear. Given scale of capital required, I don’t know how, but it is what it is.