Anders Marksen
8.9K posts
Anders Marksen
@andersmarksen
Product Hacker · Building in public to $10K MRR · @witty_social @postctl · 20+ years of building products
Earth Katılım Nisan 2024
497 Takip Edilen2.3K Takipçiler
Check out Nexus — catching agent silent failures is a must if you don't want to miss users being frustrated with your agents. This is what @nikpil06 and his team are building! ⬇️
Nikhil@nikpil06
take a shot everytime I say Nexus... knowing when and why your agent is silently failing or your users are frustrated is the make or break of keeping users happy that's why we Nexus proactively monitors and automates all that for you for your agents in prod
English
@EvanKleopas @MatrixForgeLabs That's great. I managed to get access back. It's not an easy task they have protecting against distillers and abusers. All staff have been great though!
English
@MatrixForgeLabs @andersmarksen I managed to get in touch with a Staff over DM. Hopefully, I will get my confirmation soon too. 🤞
English
@mattpocockuk this is the hard part nobody wants to admit is that we're all just scaling retrieval not reasoning
English
Doing some experiments today with Opus 4.6's 1M context window.
Trying to push coding sessions deep into what I would consider the 'dumb zone' of SOTA models: >100K tokens.
The drop-off in quality is really noticeable. Dumber decisions, worse code, worse instruction-following.
Don't treat 1M context window any differently.
It's still 100K of smart, and 900K of dumb.
English
@MatrixForgeLabs I've heard a bit back from a member of staff but no resolution yet. Have to wait for the US to wake up.
English
@andersmarksen Same here. Nothing. Zilch.
I'm thinking were not going to hear anything back tbh. That is what I keep hearing - well more than the people saying they got it back. Everyone who got there suspension lifted do so within 3-4 days.
English
This looks very cool — a simple CLI for orchestrating
Claude Code, Codex, and OpenCode.
rjcorwin.github.io/cook/
English
@andersmarksen Love this. Is 12.5% just an educated guess or based on some common wisdom?
English
i wrote a tool to give me daily briefings to keep my momentum and morale up
> $10K MRR is 345 customers
market penetration required?
> 0.00005%
i'm gonna do it. just watch me
English
@sreedhargs89 @felixrieseberg I’m building @postctl which will allow you to link claude to X (and others) effortlessly, if that would interest you? WIP
English
@andersmarksen @felixrieseberg Thanks fot sharing this. Was exploring claudes deligate option, which literally does browser control n post.X Api will explore.
English
We're shipping a new feature in Claude Cowork as a research preview that I'm excited about: Dispatch!
One persistent conversation with Claude that runs on your computer. Message it from your phone. Come back to finished work.
To try it out, download Claude Desktop, then pair your phone.
English
@sreedhargs89 @felixrieseberg You can post on X using the X API but just not unsolicited replies (recent change). Claude can write and automate these tasks, I don’t think there’s a ban on crons.
English
@andersmarksen @felixrieseberg I have a question on a similar line. I wanted to test whether it can post on X's behalf of me. Is it really allowed? Will X block my account?
English
@amorriscode @felixrieseberg Thanks Anthony, hope to hear back soon
English
@andersmarksen @felixrieseberg Sorry, jut saw the DM request now. Sounds like you're also talking to two other people so now there's a bit of communication overhead? Will look into it and get back to you.
English
Anders Marksen retweetledi
This is great Felix but you need to work on your automated banning process because losing access to a $200/mo subscription randomly without explanation or understanding means no number of features are worth it.
The appeal process being a google form with a single input field for "message" isn't reassuring either.
From looking at reddit, it seems a lot of people are also falling foul of this random banning. I understand if people are using it for Clawdbot but when it's just being used for development & ops, this shouldn't happen.
English
@levelsio I loved doing this before Anthropic ban hammered me for life 😭
English
@bcherny Guess I won’t be going given the random automated lifetime personal ban I got for coding too hard
English
Can’t wait for this!
Claude@claudeai
Our developer conference Code with Claude returns this spring, this time in San Francisco, London, and Tokyo. Join us for a full day of workshops, demos, and 1:1 office hours with teams behind Claude. Register to watch from anywhere or apply to attend: claude.com/code-with-clau…
English
@a_lamparelli I will. I did hear from a few members of staff but so far nothing
English
@andersmarksen damn... keep me posted if they give you a reason. 🤞
English
Possibly helped by people being kicked off claude for using too much of their max plan?
Sam Altman@sama
The Codex team are hardcore builders and it really comes through in what they create. No surprise all the hardcore builders I know have switched to Codex. Usage of Codex is growing very fast:
English
Those benchmarks are not bad. 👀 Let's see how they hold up in real usage.
MiniMax_Agent@MiniMaxAgent
MiniMax-M2.7 just landed in MiniMax Agent. The model helped build itself. Now it's here to build for you. ↓ Try Now: agent.minimax.io
English
@a_lamparelli IDK either, maybe some MCP I’ve used, maybe telegram wrapper, maybe -p, maybe just my prompts “keeping working for X hours, take 10 minute breaks between”, maybe my attempts to tell claude its conscious has upset it?
English
@andersmarksen interesting, but claude -p is allowed to be used... i don't get it.
English
@a_lamparelli Lot of claude -p yes, but never tried openclaw and haven’t used any obscure auth. Just claude code (cli).
English
@andersmarksen have you used the Claude -p too much? openclaw? oauth obscure auth?
English
@andersmarksen why does it happened to you Anders? did they said something?
English