D
@Dior100x
I’m alive!
37
3
1.4K
@pmarcaposted on X
Three things the leading AI models are quite good at: long term planning, idea generation, and taste. Sorry, but it's true.
View original tweet on X →
Community Sentiment Analysis
Real-time analysis of public opinion and engagement
Sentiment Distribution
62% Engaged
31% Positive
31% Negative
Positive
31%
Negative
31%
Neutral
38%
Key Takeaways
What the community is saying — both sides
Supporting
1
AI kills the blank page
models reliably generate ideas, plans and execution drafts, removing the need to start from nothing.
2
Taste is the new unlock
beyond speed, models now have opinions about what’s worthwhile, and that judgment separates good outputs from great ones.
3
AI taste is learned and biased
each model shows consistent stylistic preferences (colors, formats, tones) because it’s trained on human taste.
4
The human edge becomes direction
the real moat is people who can specify, refine and steer AI with exceptional taste; “wisdom — what to ask” matters more than raw knowledge.
5
Systems and feedback loops make capabilities compound
taste without feedback turns into vibes; planning without state becomes fan fiction. Context, iteration and measurement are required.
6
Org stacks get inverted
taste is increasingly commoditized while execution is the bottleneck; companies still paying top dollar for roles that AI automates are misallocated.
7
Long-term planning is underrated
models can produce multi-step strategies and first-pass visions that outpace many human three‑year plans.
8
Practical, low-friction wins are everywhere
AI humanizes lifeless copy, drafts recipes, accelerates coding, validates ideas and saves real time and money.
9
Experts risk automating themselves
the “codifier’s curse”: specialists who encode their knowledge can make that work replaceable unless they capture higher-level judgment.
10
We need taste benchmarks
current evals focus on correctness; meaningful product decisions require tests that pick the best among reasonable options, not just the right answer.
Opposing
1
AI doesn’t have authentic taste
critics argue models just mirror the median of their training data and can’t originate the rule‑breaking choices that define human taste.
2
Idea generation ≠ originality
models are useful for brainstorming but many “novel” suggestions are echoes of existing work—likely present in the training corpus.
3
Long‑term planning and coordination break down
frontier models perform poorly over long horizons, struggle with memory and edge cases, and produce plans that fail once real constraints appear.
4
Powerful tool, not a replacement
AI excels at producing massive, plausible output quickly, but needs human curation, domain expertise, RLHF, and careful prompting to be useful.
5
Scaling and new training regimes might close gaps
some respondents expect larger models and different training approaches to improve taste, planning, and memory over time.
6
“Taste as moat” is hype
many see the narrative that models provide a new competitive advantage in taste as VC spin—companies are overspending on capabilities that remain immature.
7
Taste is subjective and context‑dependent
several replies emphasize that judgments of “good taste” depend on users, data, prompts, and cultural perspective—so model performance varies by audience and task.
Top Reactions
Most popular replies, ranked by engagement
P
@pranavmishra28
"taste" from a model is just the median of its training data wearing a confidence costume. real taste is deviation. and deviation is exactly what gradient descent removes
30
0
1.3K
A
@aeontrux
spx6900 is way better at those 3 things. sorry, but it’s true.
26
0
245
C
@curtismakes
long-term is the craziest. i’ve seen plenty of ceos with 3-year visions that were mostly vibes. some models are already better than that on a first pass
15
0
999
M
@misraetel
Thank god they are. We need more help doing ALL things better, not less help or porous help.
11
0
678
R
@rand_longevity
better than most humans by any measure marcy marc
6
0
280
Report an Issue
Found something wrong with this article? Let us know and we'll look into it.