🍋

building, learning and sharing

[🍋 @humanwritten_]

my favourite @steipete quote was something along the lines of testing openclaw by asking it to log into the @British_Airways site and being a greater challenge than the turning test, to which I wholeheartedly agree

[🍋 @humanwritten_]

have anthropic added anti open claw rl in 4.7 or something or is the generative blackbox just doing it's thing?

[🍋 @humanwritten_]

@theo @DannybSZNNN i already anticipated they would do this and have used codex for all my analysis

[🍋 @humanwritten_]

@thsottiaux @steipete @bcherny fyi for you also, i'd love to publish this app and can rename if needed, but would rather not as i love the little wand / duster thing that i gave them

[🍋 @humanwritten_]

@thsottiaux ~20x as described :) - an old app named clawd before the entire clawdbot saga - i've held off publishing since the naming chaos - fyi @steipete - maybe this from threads adds some history to their internal naming and the chaos that ensued

[🍋 @humanwritten_]

@steipete on it

[🍋 @humanwritten_]

@CryptoFlammel @ricardo1 @bcherny yep all was fine full recovered. however the reason for the ban was never clear.

it would have been good to have some insight, as people in reddit speculate it's because of vpn's or being in multiple countries, something that many developers do all the time.. so 🤷‍♂️

[jack @jack]

everything is programming

[🍋 @humanwritten_]

worked for me :)

[am.will @LLMJunky]

If you have an Anthropic Plan, stop what you're doing right now

Go claim your free credits on your usage page. (Link in comments)

They are gifting users the same value their plan costs, which is nice.

They have finally clarified: no third party harnesses allowed on their plans.

You can still use them, but at discounted API pricing.

Glad they finally clarified this, even if I don't like it.

[am.will @LLMJunky]

If you have an Anthropic Plan, stop what you're doing right now

Go claim your free credits on your usage page. (Link in comments)

They are gifting users the same value their plan costs, which is nice.

They have finally clarified: no third party harnesses allowed on their plans.

You can still use them, but at discounted API pricing.

Glad they finally clarified this, even if I don't like it.

[Christof Salis @christofsalis]

I know Claude code has buddies, but PI has grumpy old guys judging you. Occasionally, he swears. If he could ban you for your slop, he probably would (but don't worry, he can't).

@badlogicgames any characters are purely fictional. inspired by real events

[🍋 @humanwritten_]

@anything 20k credits, happy to try it out

[🍋 @humanwritten_]

@larsencc need to create like an isEven for isPort - use 3000, if unavailable + 1, break once found - maybe avoid some magic numbers

make no mistakes..

[🍋 @humanwritten_]

@theo @trq212 i can neither confirm nor deny

[🍋 @humanwritten_]

@larsencc now i just want the agents to know the ports are in use without having to check (or needing to force checking automatically)

i was surprised this wasn't more resolved natively but i suppose until recently you weren't so likely to be running 10 totally separate projects at once..

[Nandkishor @devops_nk]

Honestly, this is the most accurate diagram I've seen.

Waterfall: You plan for 18 months and deliver exactly what nobody needs anymore.

Agile: You deliver something usable at every step, but the CEO keeps asking, "Where's the car?"

AI: You get the car on day one. It has six wheels, the doors are on backwards, and it has a rocket launcher. You spend more time making it yours than actually "building"; it's shaping. owning. verifying. That's what the best AI developers do now. They don't build. They shape and own.

[Mario Zechner @badlogicgames]

btw, @bcherny put up a bunch of prompt cache fixes for openclaw.

https://t.co/Z7wE167SPl

if you are building custom harnesses, read them. prompt caching is really not hard to understand. but almost all harnesses get it wrong. it's puzzling.

[🍋 @humanwritten_]

@BenignRadicals @bcherny see my other comment around whether for example users could send some kind of self id for an external harness and voluntarily pay for for extra usage via their main cc account and extra usage mechanism instead of a separate console account

[🍋 @humanwritten_]

@BenignRadicals @bcherny i mean someones got to develop the enforcement into the product right .. ?

[🍋 @humanwritten_]

i said this the other day but i feel that plan mode was a crutch for claude code being way to trigger happy in starting to edit code before gathering context

however, instead of just restricting edits and writing to /tmp and just putting on a leash, plan mode locks claude in the house.

i remember interactions going more like "question, task, context etc. pls don't edit code"

... *spelunking* okay done heres the analysis you asked for

... okay great lets go i've edited 10 files 4 of which were irrelevant - code rollback doesn't exist yet and you forgot to git commit before asking me not to change things, sucks to be you

it is much much better now but this is how i remember it.

i'd still love a shift + tab mode that goes into a restrict edits but not plan mode kind of thing