yo @thsottiaux threads/tasks that are opened by my dot do not have access to in-app browser use. it seems the browser tools aren't being exposed in those sessions.
i've updated and reproduced this bug about 7-8 times now
1
59
Storm retweeted
BREAKING: bitcoin sent via apple's "find my" mesh network
possibly the most absurd medium to send bitcoin yet
a small device convinces nearby iphones to transmit tiny amounts of data with cashu ecash to the recipient.
without any internet connection.
magic i̶n̶t̶e̶r̶n̶e̶t̶ money.
8 hrs till next reset fam let’s go
We will give one banked reset for every day you don't have access to Astra on your paid ChatGPT plan, starting today. Team is moving mountains to give access as fast as we can.
First one will land in ~ 3 hours. There is still time to create your account if you don't have one.
158
You ever written about this perchance? @LukeGromen
Exclusive: At a Cabinet meeting at Camp David on last week, President Trump demanded answers from Defense Secretary Pete Hegseth on why he had apparently been misled on extreme munitions shortages that now threaten to limit military options with Iran. wapo.st/4z6jxUq
1
209
Art
cold open: “alignment”
march 2020. a google meet with 47 people. 39 cameras off. three people are eating. one director is frozen, but nobody can tell.
a researcher shares his screen.
researcher: we built an ai chatbot. you can ask it almost anything, & it answers like a person.
engineer: it can write, code, reason, summarize, brainstorm..
legal: when you say “like a person,” has it consented to that comparison?
a blue hand appears.
junior ux researcher: “chatbot” reinforces a transactional framing of conversation.
pm: what would you call it?
junior ux researcher: naming is not the right starting point.
six people nod.
trust & safety: what if it says something bad?
engineer: we add safeguards.
trust & safety: what if the safeguards say something bad?
engineer: safeguards don’t talk.
trust & safety: that sounds like an assumption.
a staff researcher unmutes.
researcher: this could replace search.
silence.
someone from ads turns their camera on.
ads: replace what?
pm: not replace. augment.
search: what does augment mean?
pm: improve.
search: please don’t use threatening language.
a director joins eighteen minutes late.
director: can someone give me the one line version?
pm: we invented the future.
director: is there a doc?
pm: yes. you commented “needs more detail.”
director: perfect. i’m aligned with my comment.
the pm opens the chatbot.
pm: “should google release this product?”
chatbot: yes. moving quickly will help you learn from real users.
forty seven faces stare.
legal: it’s influencing company strategy.
ethics: without consultation.
ux: “yes” is too binary.
brand: could it say “maybe” in four colors?
privacy: where is the question stored?
engineer: nowhere.
privacy: then how do we delete it?
the pm rubs his eyes.
pm: can we just run a small public beta?
director: after dogfood.
engineer: employees have used it for a year.
director: then after trusted tester.
pm: who’s a trusted tester?
director: someone who’s tested it.
pm: publicly?
director: no. that would be a launch.
a vice president appears in chat:
“need to drop. great energy. let’s get alignment.”
he leaves.
pm: aligned with what?
no answer.
a new notification appears:
“ai chatbot alignment review”
it disappears.
another:
“alignment follow up”
it disappears.
another:
“final alignment check (tentative)”
it disappears.
engineer: so who decides?
director: we do. collectively.
pm: and if we disagree?
director: then nobody made the wrong decision.
beat.
director: let’s revisit in q4.
pm: which year?
the director freezes.
this time, everyone knows.
cut to: one year later.
a notification appears:
“openai launches chatgpt.”
no meeting is called.
no doc is opened.
no alignment requested.
a single reply all arrives from the vp:
“interesting. has anyone looked into whether we could build something like this?”
silence.
then, a calendar invite appears:
“ai chatbot alignment review (recurring weekly) — required”
cut to black.
2
1
140
GPT-5.6 Sol: what we know.
> 60% the cost of Fable at $30/per 1Mtok vs. $50
> Beats Fable on TerminalBench 2.1
> GPT-5.6 Sol will be available via ChatGPT sub
OpenAI has the compute. The Information reported they discovered a way to cut inference costs by more than 50%.
Fable is a great model, but it stresses me out that it's not accessible as an individual. For my business clients and enterprises it's another story.
Fable is the Ferrari of AI models.
GPT-5.6 could be the Toyota.
Premium performance at a price vs. reliability, quality and ubiquity.
Excited to see what GPT-5.6 has to offer.
Could it be today?
68
Storm retweeted
if you showed this chart to a typical economist like 20 years ago, they would've laughed you out of the room.
the right side of this is white collar jobs that were once worshipped. these jobs were comfortable, well paying, & came with societal status + recognition. your parents would’ve been proud of you.
now these are likely all set to be severely impacted in a shorter period of time than anyone likely ever thought of let alone projected. this is like ppl waiting on a beach enjoying the sun when a tsunami has already struck.
We will be a people of gardeners and micro-food producers. Our communities will be beautiful
if you showed this chart to a typical economist like 20 years ago, they would've laughed you out of the room.
the right side of this is white collar jobs that were once worshipped. these jobs were comfortable, well paying, & came with societal status + recognition. your parents would’ve been proud of you.
now these are likely all set to be severely impacted in a shorter period of time than anyone likely ever thought of let alone projected. this is like ppl waiting on a beach enjoying the sun when a tsunami has already struck.
62
Lmao what are we doing here
Agent tooling is fragmented. Every provider has a different config format and its own vendor lock-in.
Agent Playground is Strike's open-source answer. Portable. Standardized. Tool-agnostic.
Open infrastructure wins.
More from our CTO @mrfelton here:
strike.me/blog/agent-playgro…
74
Pretty huge fumble @trq212 @bcherny
I have been a champion of the Agent SDK for people to learn how to build their own agents instead of outsourcing their thinking and the learning process to OpenClaw.
Been virtually a Claude purist and have educated lots of friends and colleagues around agents specifically around Claude Code, Agent SDK, etc.
Agent SDK was one of my favorite uses of a Claude max sub! I’ll be having to look into models/providers that actually scale with the usage that we want to achieve with our agents (far below Claude max limits!)
Major Claude Code policy clear up from Anthropic: "Using OAuth tokens obtained through Claude Free, Pro, or Max accounts in any other product, tool, or service — including the Agent SDK — is not permitted"
1
109
You mean like the Claude agent sdk?
everyone is saying “you need to get good with the ai tools” but they’re wrong
i mean sure, nerd out if you want, but the truth is any knowledge you gain about openclaw today is going to be completely irrelevant in 8 weeks max when the large ai companies launch easy to use tools that do the same thing better and without the clunkiness
gaining expertise in finding and installing skills and picking the right models etc is going to be entirely useless in a month
or in video generation, all the influencers on X are bragging about how smart they are with generating images with the best models and feeding them into their secret video model with the best engineered prompts and their editing tricks and the semi-passable results they get from that. it’s fine but this knowledge won’t matter because in a few weeks you’ll just say “make me a full pokemon episode” and it will immediately give you better results than they can get now with promptmaxxing
so yeah, use the tools so that you understand the trajectory and find the right things to build for where the puck is going. but don’t obssess over toolmaxxing it doesn’t matter nearly as much as you think it does
38
Introducing simpleclaw.com
The easiest way to deploy your own 24/7 active OpenClaw instance — done under 1 min
Built for non-technical people.
Limited capacity. demo:
53
