191–200 of 208 posts
Earlier quoted context omitted.
Good News! The Models are done, and you can download them for free. Even if they stopped being worked on this moment, those are finished and usable right now, and won't get any worse over time.
Good News! I have a copy of a search engine from 2005, so I don't need Google!
I am actually setting it up and using it in one day. It is basically unusable. Many tools require homebrew. The setup of authentication is complicated. The default Docker setup is broken (github.com/openclaw/openclaw/issues/5559), and the cron job is not working ( https://github.com/openclaw/openclaw/issues/2323 ). I doubt any story in the comment is real.
Earlier quoted context omitted.
I think Karpathy[1] summarized why he thinks this is the case quite well (as described he was himself hyping it up a bit much, but there are some foundational reasons why it's a very interesting experiment). [1] https://x.com/karpathy/status/2017442712388309406
Karpathy is one of the biggest tech grifters of our time, so finding out that he's jumped on this grift train as well comes as no surprise. Actually, hang on... yep, to absolutely nobody's surprise, Simon Willison has also hyped this up on his blog just yesterday. The entire grift gang is here, folks.
Real user here, 7 days in. Mac Mini M4 from Costco (~$640), running Opus 4.5 via Claude Max. Using it as an autonomous employee for two actual businesses (SaaS + AV rentals in Orange County). What actually works: - Overnight autonomous work is the killer feature. Directive before bed, structured deliverables in the morning. Research reports, competitor analysis, lead lists — genuinely usable, not demos. - It found an…
Been using it daily for about two weeks now. The setup is admittedly not trivial — took me a solid afternoon to get everything configured the way I wanted. But once it's running, the value compounds fast. The key insight for me was starting small: one channel (Telegram), one provider, simple tasks. Then gradually expanding. The people who bounce off it usually try to set up everything at once. My suggestion: pick one…
I genuinely do not understand what the benefit of this tool is, over having Claude Code/Codex running on a VPS or your home machine and accessible over Tailscale. If you want to be able to interact with the CLI via common messaging platforms, that's a dozen-line integration & an API token away?...
The other non-obvious bit is operational discipline: rate limits, auditability, and being able to run it in a constrained environment (VM/container, minimal mounts, separate accounts) so a prompt-injection doesn’t turn into “oops it had my whole laptop”. A DIY script can be safer too, but most people don’t end up doing the guardrails.
That said, I agree the core idea isn’t magical—packaging is the product here, and it’s still early/buggy enough that DIY often wins if you’re already set up.
Found it actually performs well for data analysis and running automated checks (cron-style tasks). Because the reasoning layer seems reliable so far, I've hooked it up to a simulation account (paper trading) to backtest the execution in real-time.
It’s good enough to monitor the market, but I’m keeping it sandboxed until I see consistent behavior over a longer timeframe.