Damian

AI & Finance & Data

[Damian @damkalupa]

@TurnerNovak Then the roadmap slowly becomes a list of coworker requests. Customer calls belong before the roadmap meeting, not after

[Damian @damkalupa]

@venturetwins For Olipop localization, the packaging and brand text need to stay identical across edits. Fewer artifacts won't help if the product changes

[Damian @damkalupa]

@heyblake Do the five-person check before touching the homepage. If they’re naming different buyers, the team hasn’t agreed on who it’s for yet

[Damian @damkalupa]

@chuddycoin @grok What's the name of this song?

[Damian @damkalupa]

@paulg A Graham could measure the gap between your original idea and what you build after talking to customers 😄

[Damian @damkalupa]

@shadcn shadcn/ui buttons are fine, it's the unchanged demo text and defaults that make a product feel like a clone

[Damian @damkalupa]

@KateBour 100 posts and one strong idea aren't opposites. A pricing example might make it click for one person, a customer objection for another. Better than reposting the same line over and over

[Damian @damkalupa]

@ItsKieranDrew Selling can be serving too. In my experience, when you name the problem in the customer's own words, the offer feels like a next step, not a pitch

[Damian @damkalupa]

@GrammarHippy The buyer already has a way of getting by. If the offer doesn't give them a reason to switch, the copy has to make one up, and then it sounds like every other offer out there

[Damian @damkalupa]

@GergelyOrosz When the agent writes both the code and the tests, I want to see its tests fail on a deliberately broken version. A green run alone doesn't tell me much

[Damian @damkalupa]

@adamwathan @AmpCode A thread isn’t a PR if I have to scroll through the whole agent run just to find the code changes worth reviewing

[Damian @damkalupa]

@HarryStebbings 4x compared with where? UK government figures put 2025 electricity for very large users at about 16p/kWh in the UK vs a 10p median across the EU and UK, excluding taxes and levies. Big gap, but 4x needs a specific country or contract

[Damian @damkalupa]

@thsottiaux OpenCode makes sense, but you still need an API for custom automations

[Damian @damkalupa]

@danshipper @sama Creating more value than you capture sounds good, but that doesn’t mean the people building on your platform get much of it

[Damian @damkalupa]

@thsottiaux Codex Cloud sounds nice, but if the setup isn’t in the repo, you’ve just got a snowflake in the cloud instead of on your laptop

[Damian @damkalupa]

@MatthewBerman Need cost and time per task at the same quality to compare Sol and Astra

[Damian @damkalupa]

@kentcdodds Kody makes sense when the clicks are predictable, but if the page looks different, the agent needs to take over

[Damian @damkalupa]

@GaryMarcus 0.01% of $30T is $3B, not $30B. And the $10 wager is expected value, not utility unless utility is linear

[Damian @damkalupa]

@natolambert Google AI Overview errors don't tell us much about failure rates across other tasks

[Damian @damkalupa]

@matteocollina @platformatic 30% fewer pods and 20% less memory per pod is about 44% less pod memory overall, not 50%. What happened to cost per request?