@atmoio@addyvantage WRT that I wonder what the over/under on human labor costs are when it comes to marketing. If coding agents were "there" it'd be a no brainer that offloading specs to competent AIs is the "move", but as it stands... 🤷♀️
I hope you find and solution and thanks for the videos 🤠
@atmoio@addyvantage I guess for me, as someone who enjoys crafting coding solutions myself I go: "I don't even wanna do this and it doesn't work and it costs me money to do it this way wtf am I doing". But when I don't embrace it I feel the (I think manufactured) dread that Im falling behind.
@addyvantage@atmoio To me at least it's also that success with an LLM is highly task dependent even though my ability to describe the problem space in English 'feels' homogenous... this leads me to doing what works consistently 'trad coding' because i get annoyed at the inconsistency of the outputs
people giving an llm full unauthenticated access to your system are peak 2026 tech bros. the clanker can run an rm -rf just because an unread spam email told it to
you guys are running a personal assistant that has 512+ critical vulnerabilities. 2026.1.15 patch added a backdoor for telemetry and nobody even read the pull request before starring its repo
imagine your vibe coded "agi" gets hijacked the second you open a malicious tab and it gives up your aws keys. it is hilarious watching youtubers cry about openclaw and the users are willingly typing their seed phrases into it just because youtubers told them to.
if your workflow relies on a buggy react application and a mcp server, you are ngmi. openclaw executes code first and never asks for confirmation.
people are actively bypassing the new ssrf protections in version 2026.2.12 by just using ipv6 addresses. the fact that this app has 150k stars proves that developers do not care about security at all.
i refuse to use a local ai agent named molt or claw. indirect prompt injection is unsolvable by passing raw browser dom to claude.
once give it access to your apps and repos, the moment someone pushes a prompt injection commit, your helpful bot could deploy a crypto miner.
you dont need a zero day to hack openclaw, you just need to send the user a text message containing an xss payload. the bot reads the message, parses it, and instantly leaks your ssh keys.
over all that, you're comparing an mcp server with an open source operating system with almost 0 ai generated code. you must be insane!
People at AI frontier labs: “You only have X years to escape the permanent underclass.
Of course, we’ll be part of the permanent overclass. Escaping is for you to worry about.”
Also people at AI frontier labs: “Why does the general population despise us?”
@thdxr@BrianVia Dammit Dax! What an insanely based take! My relationship with you (one sided as it is) is so messy. You’re right virtually 100% of the time, but I feel this need for you to be wrong since I have such distaste for the state of AI in software pop culture. My poor lil divided self
@edandersen@AuctusMacula@unclebobmartin First off, yes it was pretty obvious to me (not that it matters lol)
Second, I think big Bob might get more traffic and more AI glazing responses than make sense to us.
Third, and of most interest… if it wasn’t recognized as sarcasm why did he respond at all 🤔 💀
Mysterious