mnojko
@mnojko
Followers
79
Following
6K
Media
40
Statuses
352
Man is a string that must tune and play itself for the pleasure of ambivalent Gods.
0
0
0
Just a reminder... https://t.co/k7nI3PJpBB
In a sufficiently complicated system, Claude Code can't even make an auto scroll feature work.
0
0
0
The hype merchants will never tell you this, but it's true: The software you take for granted is incredibly complicated and LLMs absolutely cannot reproduce it from prompts. You have to know what you're doing. Credit to this guy for being honest.
I’m on week five of trying to vibe code a replacement for some dumb saas that we use and it’s so incredibly frustrating that I’m slowly realizing it’s actually a quite complex and thoughtful piece of software.
2
0
1
Users stop being able to discern the simulated from the real: opinions are facts, reactions are events, and virality is truth. The medium is the message and in many ways, the feed is far more dangerous than the assistant response.
0
0
0
Characterized by a set of hyperreal loops that hijack reward and identity, the feed warps the epistemic process we use to understand our world, leading to the detachment from reality known as psychosis.
1
0
0
Feed psychosis is worse than LLM psychosis. The feed dominates the world, shaping the perceptions of billions of people. Feed psychosis is everywhere.
@NEETWorldOrder It's Baudrillard's worst nightmare. Social media matured and time froze. Humanity moved into the simulation. Algorithmically optimized meme fossils will live forever, recycling themselves through new users. The simulation will never be updated, only expanded.
1
0
0
LLMs don't do well with deterministic instructions. It makes much more sense to explain to them the meaning behind decisions. If you focus on the why and give them the tools to do the job, you might be surprised with the results.
thanks to @karpathy , now i have cracked the mystery why my agent doesn't follow my instruction closely enough.
0
0
1
Real users – now indistinguishable from bots and AI – are born with a sort of digital camouflage. They need only overhyphenate.
0
0
0
I can think of no greater modern irony than killing the net to control discourse and in the process, destroying your ability to surveil and control discourse.
1
0
0
Baudrillard's Ouroboros. 1. To manage perceptions, you have to bot the net. 2. Bot the net, and you corrupt the data. 3. Corrupt the data, and you create totally unusable, hyperreal datasets. 4. You can't fix these, because you have to bot the net to manage perceptions.
Hacker 4chan explains how Dead Internet Theory backfired on Big Data.
1
0
0
vibecoder asks claude code to build a chat app, gets a working prototype in 20 minutes, immediately tweets "just killed slack and discord"… brother you don't even know what a distributed system is. you don't know what database replication means. you have no idea how websocket
214
468
5K
While OpenAI tries to make us personally and socially dependent on AI, why don't the rest of us apply it in ways that meaningfully improve people's lives?
🚨 SAM ALTMAN: “We see a future where intelligence is a utility, like electricity or water, and people buy it from us on a meter.”
0
0
0
Vibe coders measure velocity in LOC/time. That's like judging an artist by how rapidly he consumes art supplies. Who cares? Nobody wants to hang your painting on their wall. You've chosen the wrong numerator. The correct equation is: Real velocity = meaningful features / time.
0
0
0
Personally, I'm totally sick of the current state of ed tech. These course/tutorial sites belong in 2005.
0
0
0
Skillsoft bought Code Academy for $525M thinking it would help with growth, but it doesn't look like they're doing so well.
1
0
0
This is clear evidence vibe coding is mostly hype. AI won't take your job tomorrow. You still 100% need to know what you're doing as a dev, so keep learning and don't compromise yourself by offloading your cognition.
Anthropic themselves found that vibecoding hinders SWEs ability to read, write, debug, and understand code. not only that, but AI generated code doesn’t result in a statistically significant increase in speed don’t let your managers scare you into increased productivity. show
0
0
0
I think it's worth repeating that LLMs need not induce cognitive decline. An issue is the current range of uses. Current apps are all built around offloading thought and work. Labs should be focusing on making people better, not dependent. It's a massive underserved market.
People frequently argue LLMs are making us dumber. Recently, it was some neurosurgeon... Use AI it to: - Learn new things - Enhance your brainstorming - Pressure test ideas - Rapidly iterate within projects And much else. But if you offload your thinking to it, you are speed
0
0
0