Sixteen years building products taught me one thing: Scrum feels smooth only when teams treat it as a craft, not a checklist.
Daily time belongs to developers.
Technical calls stay with the people who write the code.
Backlogs breathe when refactoring and testing live inside them.
Story points thrive as complexity, not calendars.
Sprints shine when acceptance is crystal clear.
Burndowns help only when they stay inside the team.
And a demo with something genuinely cool lights up the whole room.
Scrum becomes fun the moment it serves the builders
Funny thing: I hear claims “we follow Scrum”or “I know Scrum then the real process looks nothing like it.
@hosseeb abstract risk never moves anyone. a real breach does. people tuned out years of safety essays, then one hack hits close to home and the whole mood flips. security incidents will do more for Ai caution than any manifesto imo
@TheStalwart the fyp inbox is gonna optimize for what you'll open. problem is your accountant email is never as clickable as some newsletter, so the boring stuff that actually matters gets buried. same trap as every algo feed
@const_reborn 110B fully permissionless is nuts. syncing gradients across nodes you don't control is usually where these runs quietly fall apart. how the comms overhead is holding up at 10x scale?
the biggest thing i got wrong selling to therapists: i thought they'd care about features. they care whether you understand their day. every good conversation i've had started nowhere near the product.
i'm still pre-raise, so treat this as a read and not a war story. but it already changed how i talk about what i'm building. i stopped defending the floor and started describing the ceiling.
getting ready to raise a small pre-seed and the thing that reframed everything for me: most vcs are not trying to fund a good business. their fund only works if one bet returns the whole thing.
which changes what you pitch. a safe path to a real business reads as a no. what lands is the world where this gets enormous, and why i'm stubborn enough to be the one chasing it.
@suchenzang that 90% in 3-6 months call aged like milk. ai writes a big chunk of my code now, but "controlling the machine gods" is way too flattering. mostly i babysit a junior who never sleeps and occasionally nukes prod. someone still has to say what to build
@leopardracer the switching cost is the real tax. i bounce between claude code and codex and every handoff the new one rereads half the repo just to catch up. shared semantic layer makes sense, my only worry is agent B inheriting agent A's wrong assumptions
@alaphati_t the review step is the one everyone skips. had AI hand me a clean looking function last week that quietly swallowed errors in a try/except. read totally fine, wouldve been a nightmare in prod. you really do have to read what it gives you
@djcows lol put one next to every standup and we're energy independent by Q3. jokes aside the whole AI race quietly turned into a power race. everyone fights over GPUs but the actual wall is just watts
code being done and the feature being done are weeks apart. the last 10 percent is error states, empty states, what happens when the network drops mid-request. none of it shows up in the demo. all of it is the actual work.
cost-plus pricing feels safe because you can defend every dollar. but you're only defending it to yourself. the person across the table is running completely different math.
@StockSavvyShay the demand side is obvious now. the part nobody says out loud is the grid physically can't keep up. interconnect queues run years and there's a real transformer shortage. software can optimize the grid all it wants, you can't automate a substation that doesn't exist yet