@udiWertheimer The family sharing plus YouTube Premium does make the math look good on paper. Still, I keep ChatGPT around for the stuff Gemini fumbles, so for me it ends up being both, not either.
@ivanainai The per-build price gap is real but it is not the deciding factor, the cheapest build still loses if it needs a second pass. What matters is which model lands a clean build the first time, that is where thirty-two cents beats fifty.
@NielsRogge@YouTube@googlephotos The funny part is Google competing with Anthropic while owning a chunk of it, Neon catching up to Opus is a win either way for them. With search, YouTube, and TPUs on top, they are hedged no matter which model developers pick.
@MiaAI_lab Generating 100 files is the easy part, keeping all 100 visually distinct is the real test. Most models start repeating layouts after the first dozen, so if Neon held that variety all the way through, that says more about its taste than its speed.
@MattNiessner The tools are already good enough to draft and to review, the trust just hasn't caught up on either side. The fix isn't banning LLMs, it's disclosing where they were used and judging the work itself.
@dabit3 Outcome engineering is the right framing for 2027. Orchestration got agents running in parallel, but the bottleneck now is defining what done looks like and verifying it, not spinning up more agents.
@Ionkosm Distribution. ChatGPT is already where everyone works — inside Office, inside iOS, inside every workflow — while Claude is still the better tool waiting for people to come find it.
@jasondeanlee That matches what I found too. Gemini burns through usage fastest and the math slips are real. I kept Claude for writing and slides as well, GPT for the daily grind.