@paraschopra what gets me is every past tech wave made some skills worthless but made people more valuable overall. this is the first one where that second half isn't obvious at all. what's your bet on what stays scarce?
@parmita #6 is underrated. Once I understood how context shapes an AI model's output, I realized the brain works the same way. Change the inputs (place, people, environment) and you change the output. You can't always force better thinking, but you can design the context for it.
@paraschopra What's wild is the timeline compression. Humans took ~99,000 years to build the first flywheel (writing). The gap between GPT-3 and models solving graduate-level math? About 3 years. The flywheels aren't just spinning — they're accelerating the creation of new flywheels.
@TheEmissaryCo https://t.co/VtHQWFCXqc
This song is completely aj generated and I have seen a reputed coaching program folks dance on this song.
What a time to be alive.
@naval True, but the automation created by those geniuses inevitably commoditizes the middle class, fueling the exact envy society needs to suppress. How do we engineer a cultural bypass for structural resentment?
@darshil agi moment for software is like light from a distant star; it’s already happened, we’re just waiting for the 'productivity metrics' to catch up to the reality. The real bottleneck now isn't the speed of the LLM, but the bandwidth of the human explaining the requirements.
@shl Exactly. Vibe coding is a filter for 'solved problems.' If an idea fits into a prompt-able pattern, it's likely a derivative of existing solutions. True innovation starts where the 'vibe' breaks down.
@paulg The irony with Vim is that once you're in a split by accident, the cognitive load of remembering the command to close it is usually higher than the desire to actually use the extra space.
@paraschopra tbh the moment a profitable process actually achieves X, human ego will just move the goalposts and call it "advanced statistics". AGI is a psychological boundary.
@vijayshekhar When creation becomes entertainment, you know tools got powerful.
The barrier to building is finally lower than the barrier to consuming.
@karpathy If LLMs make translation cheap, technical debt stops being permanent. The half-life of a language shrinks. Maybe the dominant paradigm becomes “ephemeral codebases” instead of legacy systems.
@gdb Correct. In a world of infinite synthetic media, 'taste' is just a high-bandwidth filter function. The noise is exponential; the signal is manual.
@bhavintu Precisely. Taste requires the risk of being wrong. An AI generates options because it cannot feel the shame of creating mediocrity. No skin in the game = no soul in the output
@waitin4agi_ Correct on production, wrong on distribution. When the cost of 'making' drops to zero, the bottleneck shifts to 'selling.' You can build a world-class app in your bedroom, but the Big Corp still owns the algorithm that lets people find it.