Aw Khai Sheng

@basedjensen @psantloki and u dont think they arrived at the existing pricing from experimenting?

@psantloki Only a finance bro would think they need to experiment on packaging like this especially when the feature has been what has been generating revenue and growth

Anthropic bros I don't know who this guy is but this guy single handedly just did massive brand reputation harm that is going to show up as percentage of loss on arr. You all are massively fumbling the bag

@lift_deadalus @eringriffith @tenobrus aside from no AI slop patterns, i was biased by this typo

@eringriffith @tenobrus It’s most probably generated with Claude. I worked on a huge report with Claude and it was very similar to the substack. The substack is pretty good nonetheless

A detailed and brutal look at the tactics of buzzy AI compliance startup Delve "Delve built a machine designed to make clients complicit without their knowledge, to manufacture plausible deniability while producing exactly the opposite." substack.com/home/post/p-19…

asked claude to build off the existing /insights command and make me a Claude Code Wrapped! purely with a skill. ngl im skeptical about some numbers hmmmm that Day of Week chart is cool tho!! it even profiled me as well!!

@RobertJBye Would be great if user level skills could be used in claude code mobile !

We’re making the Claude mobile app even better, so please share your feedback! What annoys you about it? What bugs are you seeing? What features are missing?

@0xDevShah my agent would start scrolling X half the time

the most valuable openclaw wrapper would be an agent that watches you work for a week, learns everything you do, then does 90% of your job autonomously. i see 100s of new clawskills on github every day, but that is bottom up. a true top-down would be your agent continually learning from your actions. that's the next unicorn. who's building this?

@JoshDaws @mattpocockuk yes!! which is why i always reply this on every 1st draft of the plan: "because i intend to clear context before executing, pls include more of the WHY in this plan thanks"

@mattpocockuk Counter argument. Plans are lossy. They lose info. Especially if you clear context when you approve. Make sure everything you’ve talked about ends up in the plan. That nothing is just assumed from the conversation.

I've stopped reading the plans that Claude creates. Frederick P. Brooks (Mythical Man Month guy) talks about a 'design concept' which is shared between all designers, but separate from all assets. I can tell from the quality of the conversation before the plan whether me and the AI share the same 'design concept'. I often get it to grill me for a long time, far past its own instincts. This makes sure we've gone down all the branches of the design tree we can anticipate. The plan is then just a compacted version of the conversation. I don't need to read it.

added thinking effort to my statusline! What does yours look like?

@MatthewBerman @openclaw is the rag done using qmd?

New use cases for @openclaw: * Self-evolving personal CRM - tracks relationship details of people I email with * Food tracker (been having stomach issues) - tracks my meals by text or photos, will run analysis later * RAG knowledge base - ingests articles, videos, tweets, that i can later recall for videos

@shiri_shh How will it be different from Openclaw on my phone?

Rabbit R1 should just pivot and start selling its device with OpenClaw pre-installed

why is vibecoding still used as a derogatory term Are people still literally typing code??

@ravenxone @mignano That last idea is useful but unless granola is connected with all other enterprise data the output will be very lacking I want maximum control to create the tasks myself after connecting granola with other sources. I alr do this and even the AI output is not 100% there yet

@mignano I like the strategy! I’m just curious, could it negatively impact granola given meeting notes is a commodity and there’s little to no pricing power? MCP leads to further commoditisation Would it have been better if they instead turned meetings into tasks and then actions?

The rulebook for building a successful software business is being completely rewritten. 5 years ago, a feature like this would have never shipped, because the entire business of platform building was to keep users inside of your product's pixels at all costs. Get as many users in your product as you can, and retain as many as possible. Now, what matters most is helping users generate the most unique context possible, and giving them the power and control to take that anywhere they want to want to go. Feedback on this launch so far suggests Granola's context is indeed extremely unique...

@corsaren "Pre and post-compaction hooks to improve coherence over long tasks" this sounds interesting! what happens in the hook?

I’ve finally got Claude Code automating my job for me in a coherent way: - Dedicated work folder—no permission to edit files outside but can copy files in - Dangerously-skip-permissions + safety hooks - Enhanced memory system + work tracker skills - Pre and post-compaction hooks to improve coherence over long tasks - Daily working folders - complete chaos, you do you buddy - Weekly cleanup & review (scheduled on my calendar + cron job); add new skills, scripts, structure files, etc. All of this is still janky as hell, but I think it’s working…

i'm finding myself writing this more and more in Plan Mode its now permanently on my clipboard: "because i intend to clear context before executing, pls include more of the WHY in this plan thanks" i dont know whats a better solution

@GaelBreton It’s Claude code but for group chats

Every Moltbot / Openclaw demo I've seen shows the same basic stuff. Calendar invites. Email monitoring. Support tickets. Morning summaries. Listened to Greg Eisenberg's whole podcast on use cases. Nothing was new. Anything you could do in Moltbot, you could do in n8n. Or Claude Code. We've been sharing these exact workflows for two years. Show me something Claude Code couldn't do before and I'll be convinced.

AGI: when claude admits he's an overthinking mess

@Nicolascole77 As long as there’s competition I think the prices will remain low Cowork might one day be default jus like MS word

a terrifying realization I had tonight playing with Claude Cowork is that, right now, we have the luxury of learning how to use AI "for free." But eventually, all these companies are gonna need to become profitable, and AI usage costs are going to go up 10x-100x. good luck to the everyday person trying to learn how to use AI when it costs $2k or $20k per month which means, if you think the skill/wealth divide is wide now... just wait 5 years. It's going to be cataclysmic

@mokshmsharma @jasonfried but how do u know you’re elite even u haven’t launched

@jasonfried Building in public is a great way to optimize for likes, but a terrible way to optimize for mastery. Get good in the dark. Launch when you’re elite.

100%. Early obscurity is your friend. You're going to do all sorts of shit wrong, be an idiot in all sorts of ways, and demonstrate how little you really know about what you're about do to. You don't want to do this in the spotlight. Take the pressure off. Preserve your ego just a little — you'll definitely need it later.

@nicbarkeragain can't be a worse programmer if u dont know programming to begin with 😀

You have to admit these two widely accepted beliefs are funny: - you must use AI programming tools to avoid being left behind - using AI programming tools makes you a worse programmer algebraic substitution: "to avoid being left behind, you must become a worse programmer"

i’m building Jira for couples 💕 Hot take: you don’t fight about chores cos of laziness. You just need a system TestFlight is live now!!

@mattpocockuk Is this bait But ngl I actl agree, subagents are glorified for small tasks. Your main agent might even be losing context if it keeps delegating it away

Overusing subagents will cost you tokens If the main agent already has enough context to complete the task, just use the main agent And if you need to specialise it, use a skill