Everything I publish, in the order I published it: dispatches from Twitter, videos from YouTube and notes from my vault, archived here in full — text, images, video and links — so they outlive the platforms.
What if your cloud was actually yours? 🤔☁ With Storage Share, you get real control, easy collaboration, affordable storage, and even more to explore! Find out more in our latest #TkkBits. t.co/oldid7nUbP
That was my first thought as well. At the same time, what if those companies who already have built their on-prem agents (often using Anthropic API anyways) migrate them to Managed Agents instead?
I mean, yes, the customers they're serving are - in theory - just one Anthropic platform account away of building their own agents but I still want to see this happen for real.
Like someone saying: "You know what? I know nothing about filing taxes for my business. I will NOT use a traditional OR AI-first Saas but build it myself instead."
At least I would not. For the same reason I decided to pay for Wispr Flow a while back instead of vibe-coding it over a weekend.
12. Sixtyfour (@sixtyfourai) so agents can search for people and companies.
13. Exa (@ExaAILabs): so agents can search the web (Google doesn’t work for agents)
If you stitch all of these together, you get a digital coworker that looks more human than AI.
@shivsakhuja@agentmail@tryagentphone@grok can you make a comprehensive list of the agent-first apps and services mentioned in the parent and across the comments and sub-comments here?
@AmazonKindle I love, love, love every single Kindle I've ever bought.
And speed absolutely matters:
My latest one, an early 2025 Kindle Paperwhite, is so fast that I wound up reading Pratchett's 42-volume Discworld series in less than 12 months.
And yet, since buying my first Kindle, I still have one single feature request:
After I've continued reading a book on my Kindle
And I've then locked the screen
When I look at the lock screen
I want to see the current page that I had been reading
So that whenever I glance at my locked Kindle
I can quickly read a sentence or two of what I was reading
And pull myself back into continue reading that book
And keep my reading momentum
And finish more books sooner.
The "whenever I glance at my Kindle" happens relatively often and neither the book cover nor the default graphics help me keep momentum the way the "open last page" would.
@SpringStreetNYC right now, I don't find a traditional repl very useful anymore, but that could all change (since things are changing lightning fast) and it could become useful again
Thank you!
I got the idea of nanoclaw having/being a REPL as I was reading your "Land Of Lisp" chapter on autonomously evolving the little "M" creatures.
I loved how, in your example, the biosphere's snapshots of 100 and 5.000.000 iterations kind of looked the same but then, looking at the "biomodalities" after those millions of rounds, we discover how two completely different species evolved!
This was SOOOOO cool. I felt like I was watching evolution happening live under a microscope.
My wife after casually teaching herself @claudeai Claude Cowork since Tuesday:
"I couldn't fall asleep last night because I realized I could almost completely let Claude Cowork do my taxes this year.
All I need to do is X, Y and then Z. Then send everything to the advisory.
I could be done with the whole thing in half a day."
As someone who builds software for our family office's tax advisory all day:
Just a moment ago, I wrote about switching to API-usage, "playing fair and by the rules" and still feeling conflicted about it:
And long story short, I personally feel better about myself by playing by (what I believe to be) fair rules.
And it's funny because does this really matter?
Do I, as the subscriber, really have to take into account how my actions affect the "feelings" of a "gazillion-dollar company" that I'm paying for a service? Is the "all you can eat" buffet analogy even true?
Long response but, I have mixed emtions about a lot of it, including my own decisions.
For me, a bunch of stuff came together at the same time:
Disclaimer:
a.) I was still using OpenClaw at this point (but the same happened with nanoclaw)
b.) I'm not affiliated with Anthropic in any way other than being a subscriber
During setup, I saw the option to use Claude Code. I figured... "Hey, I got the 90€ Max subscription and in one year, I've never reached its token limits. So, how bad could it be?"
So I chose to use my subscription auth option over plopping in an Anthropic API key. I'm thinking nothing of it. And then, you know, used it for a couple of days quiet heavily.
Shock 1: I looked at the stats and saw: I had burned through 18.4M tokens in one day. Eighteen. Million. Tokens. For basically configuring an agent to use WhatsApp and a cron job? See t.co/zs6VMJexpc for details.
I was like "wat."
Then I saw the "Anthropic officially bans using subscription auth for third party use" thread on HN. Somehow, I found TZubiri's "all you can eat buffet" analogy the most surprising:
A similar scenario would be if you go to an all you can eat buffet, you pay for a single person, but then you actually unload an army of little clones that start eating the whole buffet. Technically it was an all you can eat buffet and you paid the price right? Well no, come on, don't play dumb. - t.co/WMshusbl7n
Shock 2: Oh, that 90€/mo. plan (that has completely changed how I work) is heavily subsidized by Anthropic assuming fair-use.
Shock 3: And oh, without noticing, I had started carting off plate after plate after plate from the "all you can eat buffet" that assumes fair-use.
Shock 4: And ohhhhh shit, they could have unplugged my subscription because of all that.
All together, this threw me off and I tried to process it here: t.co/RJsaawzofc
So. For me it was all of these coincidences that really made me think about my part in the fair-use and decided that I would...
a.) rather keep my subscription intact
b.) start using the actual API
c.) keep an eye on cost
d.) make a conscious decision of cost vs. value a few weeks later
And long story short, I personally feel better about myself by playing by (what I believe to be) fair rules.
And it's funny because does this really matter?
Do I, as the subscriber, really have to take into account how my actions affect the "feelings" of a "gazillion-dollar company" that I'm paying for a service?
Is the "all you can eat" buffet analogy even true?
I don't know and I'm still conflicted about it all.
What I do know is:
That I'm paying for my "fair-use" API-based nanoclaw assistant ~$2 per day which feels completely above board.
It costs an extra $10 (one-time) to add a new skill or improve the existing ones every few weeks. Again, above board.
I feel like I'm doing "the right thing" but I could totally be wrong, too!
I mentioned your "all you can eat" buffet comment from HN (t.co/WMshusbl7n) in a discussion about using subscription vs. API.
"A similar scenario would be if you go to an all you can eat buffet, you pay for a single person, but then you actually unload an army of little clones that start eating the whole buffet. Technically it was an all you can eat buffet and you paid the price right? Well no, come on, don't play dumb."
Interestingly, your HN comment seems to have gotten downvoted but I feel like you're on to something.
Not sure if you can see this, it's inside the nanoclaw group but, yeah, FYI:
Long response but, I have mixed emtions about a lot of it, including my own decisions.
For me, a bunch of stuff came together at the same time:
Disclaimer:
a.) I was still using OpenClaw at this point (but the same happened with nanoclaw)
b.) I'm not affiliated with Anthropic in any way other than being a subscriber
During setup, I saw the option to use Claude Code. I figured... "Hey, I got the 90€ Max subscription and in one year, I've never reached its token limits. So, how bad could it be?"
So I chose to use my subscription auth option over plopping in an Anthropic API key. I'm thinking nothing of it. And then, you know, used it for a couple of days quiet heavily.
Shock 1: I looked at the stats and saw: I had burned through 18.4M tokens in one day. Eighteen. Million. Tokens. For basically configuring an agent to use WhatsApp and a cron job? See t.co/zs6VMJexpc for details.
I was like "wat."
Then I saw the "Anthropic officially bans using subscription auth for third party use" thread on HN. Somehow, I found TZubiri's "all you can eat buffet" analogy the most surprising:
A similar scenario would be if you go to an all you can eat buffet, you pay for a single person, but then you actually unload an army of little clones that start eating the whole buffet. Technically it was an all you can eat buffet and you paid the price right? Well no, come on, don't play dumb. - t.co/WMshusbl7n
Shock 2: Oh, that 90€/mo. plan (that has completely changed how I work) is heavily subsidized by Anthropic assuming fair-use.
Shock 3: And oh, without noticing, I had started carting off plate after plate after plate from the "all you can eat buffet" that assumes fair-use.
Shock 4: And ohhhhh shit, they could have unplugged my subscription because of all that.
All together, this threw me off and I tried to process it here: t.co/RJsaawzofc
So. For me it was all of these coincidences that really made me think about my part in the fair-use and decided that I would...
a.) rather keep my subscription intact
b.) start using the actual API
c.) keep an eye on cost
d.) make a conscious decision of cost vs. value a few weeks later
And long story short, I personally feel better about myself by playing by (what I believe to be) fair rules.
And it's funny because does this really matter?
Do I, as the subscriber, really have to take into account how my actions affect the "feelings" of a "gazillion-dollar company" that I'm paying for a service?
Is the "all you can eat" buffet analogy even true?
I don't know and I'm still conflicted about it all.
What I do know is:
That I'm paying for my "fair-use" API-based nanoclaw assistant ~$2 per day which feels completely above board.
It costs an extra $10 (one-time) to add a new skill or improve the existing ones every few weeks. Again, above board.
I feel like I'm doing "the right thing" but I could totally be wrong, too!
▶ Watch on Twitter
Claude PlatformAgent quickstart | Claude PlatformBuild on the Claude Platform↗ platform.claude.com
