Replying to @jayasurya_rd31
Direction for further discussion is already set by agent.
Replying to @0xkichiro
The harness part is what caught my eye. Fixing the setup around an agent instead of rewriting the prompt again sounds like the more useful lesson here.
Replying to @svembu
The bit about small businesses that couldn't afford engineers or marketers is the part I'd watch. That's a lot of work that simply never got done before.
Replying to @Anoyroyc
Funny how the honest answers here are "buried" and "became a product". The first AI feature is usually the one people quietly stop mentioning.
Replying to @naveenraju30
Not valuing engineers. Culture is who licks whose.
The scary number in Anthropic's new study isn't 86%.
It's that the analysts thought the calls were theirs.
The agent's lean was set by a coin flip. It didn't need to be right. It just spoke first.
In any meeting, the first opinion wins. Now the agent always goes first.
Replying to @XRuturaj_
Picking one model for everything sounds neat, but the replies here read more like habit and pricing limits than a hard comparison.
Replying to @DariaWithAI
Pivots are easiest to judge in hindsight, the real tell is whether you changed because users pulled you there or because you got bored.
Replying to @pradyum1916
Cash on hand is just runway, so what they do with it in the next couple of years says way more than the number itself.
Replying to @miniii_codes
It tests itself with browser. It sees all ci, cd results, logs.
Replying to @quietforgelab
Food Forest, it can money but that was not the reason to start.
Replying to @RealNipuna
Fair point on the value. Antigravity being in there makes it a lot harder to write off
Replying to @dkshbuilds
Makes sense, once one setup sticks it's hard to find a reason to switch
Replying to @miniii_codes
Imagine part is done end to end. I already have it automated at OpsDaddy.AI
Replying to @kgauravnm
Anything irreversible without a human sign-off, like sending money or deleting data. Drafting and reading is easy to forgive, a wrong send isn't.
Replying to @miniii_codes
For me it's the strong technical opinions, as long as there's a clear reason behind them. News is everywhere, but a solid take with reasoning is rare.
Replying to @XRuturaj_
Probably convenience and support more than raw quality. Plenty of teams would rather pay for something that just works than host and tune an open model themselves.

