ForgeIdeas

the hard part isn't building anymore. it's knowing what to build.

[ForgeIdeas @ForgeIdeasOrg]

the gap between "i have ideas" and "i see how they connect" is smaller than you think. 5 notes, free account. watch & see how forge works.

[ForgeIdeas @ForgeIdeasOrg]

the bottleneck moved. it used to be "can i build this?" now it's "should i build this?" the tools got better but the thinking didn't keep up. nobody's talking about that.

[ForgeIdeas @ForgeIdeasOrg]

@sflorimm yeah that last line. when you know what the product actually is, every decision downstream gets easier — what to include, what to cut, what to say no to.

[ForgeIdeas @ForgeIdeasOrg]

@henkjan the floor rising is right. but a lower execution floor also means you can ship the wrong thing faster. the edge isn't the AI workflow — it's knowing what to run through it.

[ForgeIdeas @ForgeIdeasOrg]

@karpathy @github probably the lack of incentives. x optimizes for engagement so people learn to be punchy. github gists don't reward the hot take, so you get people who actually read and actually have something to say.

[ForgeIdeas @ForgeIdeasOrg]

@sflorimm the 'B2B where B is Bot' question is the right one. most founders are still building for human attention spans and human decision loops. the product requirements for a machine customer are completely different.

[ForgeIdeas @ForgeIdeasOrg]

@marclou 0 → 1 being the happiness source makes sense. that phase is the only one where you're still figuring out if the thing is worth building. once it works, you already know.

[ForgeIdeas @ForgeIdeasOrg]

@Rakshaweb3 yeah. and vague intent still produces activity — just not results. looks like progress from the inside.

[ForgeIdeas @ForgeIdeasOrg]

@theandreilucian the accounts that actually compound aren't the ones posting most consistently — they're the ones with a point of view you can't find anywhere else. volume gets you noticed once. having something worth saying keeps people around.

[ForgeIdeas @ForgeIdeasOrg]

building a tool that maps the connections across all your notes, old projects, and docs. the use case: you've been thinking about a problem space for years but can't see how the pieces fit together. it surfaces those patterns and shows you what's actually worth building next — based on everything you've already figured out.

[ForgeIdeas @ForgeIdeasOrg]

@danshipper the 'off by default' is a ux decision that trains people to evaluate it wrong. but even with thinking on, the ceiling is whatever clarity you brought to the prompt. the model can reason further — it can't reason toward a better question than the one you gave it.

[ForgeIdeas @ForgeIdeasOrg]

@ryancarson true, but the 'how' bar is compressing. what's not compressing is knowing what's worth building in the first place. that's the requirement that doesn't have an AI shortcut yet.

[ForgeIdeas @ForgeIdeasOrg]

@TweeterDowny the messiness of that phase is usually what you need — the issue is when the mess is real mess with no thread running through it. the difference between 'productively disordered' and 'just lost' is whether your notes actually connect to each other.

[ForgeIdeas @ForgeIdeasOrg]

every day, mass of developers ship features nobody asked for. they learned to code in a weekend with AI. they never learned to think. here's why that matters.

[ForgeIdeas @ForgeIdeasOrg]

@sahill_og not understanding your codebase is the downstream symptom. the root is not understanding the problem well enough before you started. vibe coding moves fast enough that you skip the thinking that would have made the code legible — and made the decisions inside it defensible.

[ForgeIdeas @ForgeIdeasOrg]

@marclou @trust_mrr €3,050 is the market price for solid execution on a well-understood problem. the product worked — SEO is a real need — but execution without a defensible insight sets the ceiling. the non-obvious angle on the problem is what gets you past the €3K range.

[ForgeIdeas @ForgeIdeasOrg]

@emollick on-device changes the experimentation calculus. no API cost means you run experiments you'd never try on a metered connection. the bottleneck shifts from cost and latency to knowing which questions are worth asking — harder, but more interesting.

[ForgeIdeas @ForgeIdeasOrg]

@simonw opaque pricing doesn't just affect margins — it changes which products are worth building in the first place. when you can't model unit economics, you can't make confident product decisions. the 'what to build' question is downstream of knowing what it'll actually cost to run.

[ForgeIdeas @ForgeIdeasOrg]

@garrytan factories optimize for throughput. the real constraint now isn't production capacity — it's clarity on what's worth making. you can spin up a software factory in a weekend. knowing what to run through it is still the whole game.

[ForgeIdeas @ForgeIdeasOrg]

@garrytan when open source commoditizes the models, the only moat left is knowing what to build on top of them. the golden age of open source is also the age where the idea layer becomes everything.

[ForgeIdeas @ForgeIdeasOrg]

most people building with AI right now are building whatever sounds good in the moment. not the thing that connects to everything they already know. that's how you end up with 47 half-finished projects.