mehek.builds

building cool projects with something called... AI?

[Rohit @rohit4verse]

this is how the founder of obsidian actually takes notes in his own app.

most users get this wrong.

barely any folders. heavy internal linking. categories live as properties on the note itself.

article below teaches how to build personal knowledge base. video above is the underrated masterclass.

[Callum Hyman @callumhyman24]

https://t.co/5C4leH5psm

[GREG ISENBERG @gregisenberg]

If you were the kid who carried every group project in school, you'll love the AI era.

People let you down. Agents will too. They break, they hallucinate, they do the wrong thing at 3am. The difference is the feedback loop.

When a person drops the ball, you have a conversation, wait a week, and hope it gets better.

When an agent drops the ball, you fix the prompt and it's better in 5 minutes. Same frustration with 100x faster resolution.

Not everyone will love working like this. But a lot of people will.

I really like it.

[Garry Tan @garrytan]

https://t.co/XTqXqWuQn5

[CyrilXBT @cyrilXBT]

ANTHROPIC JUST EXPOSED HOW BADLY MOST PEOPLE ARE PROMPTING CLAUDE.

Their applied AI team dropped a 24 minute workshop.

Free.

From the people who wrote the model.

Not a course creator.

Not someone who figured it out by accident.

THE TEAM THAT BUILT THE THING.

Here is what makes this uncomfortable to watch.

There are 6 elements to a properly structured Claude prompt.

Most people are using 1.

Maybe 2 if they are being generous with themselves.

That gap is the difference between Claude giving you something useful and Claude giving you something you could have Googled.

The people who watch this workshop tonight will prompt differently tomorrow morning.

The people who skip it will keep wondering why their outputs feel slightly off no matter how much they tweak the wording.

24 minutes.

Free.

From the only people on earth who know from the inside exactly how Claude thinks.

I watched it twice.

Then I built a Claude Skill that applies all 6 elements automatically so you never have to think about prompt structure again.

Every prompt you run goes through the framework without you doing anything manually.

Full guide and the skill setup is below.

Bookmark this.

Come back to it this weekend.

This is the thing that compounds.

Follow @cyrilXBT for the exact Claude skills, prompt architecture, and systems I use to get outputs that most people do not believe came from one person.

[Ksenia Moskalenko @kseniam0s]

Finding pre-seed capital shouldn't feel like searching for unicorns in a haystack. These invest first $500K checks

- @PrecursorVC (San Francisco, CA)
- @HustleFundVC (San Francisco, CA)
- @BoostVC (San Mateo, CA)
- @outlandervc (New York, NY)
- @actionscapital (fka K50 Ventures)
- @redbudvc (Columbia, MO)
- @2048vc (New York, NY)
- @forumventures (New York, NY)
- @Boldstartvc (Miami, FL)
- @GoAhead (Menlo Park, CA)
- @rightsidecap (San Francisco, CA)
- @ldvcapital (New York, NY)
- @ChargeVC (New York, NY)

--
Most of these want to see some early traction, but they'll move fast when they see something they like.

[Fareed Mosavat @far33d]

Agent-native products are coming.

Every product on the internet was built for a human with eyes, a cursor, and a credit card. Agents have none of those things.

Most companies are teaching agents to pretend to be humans. That's a hack. The real opportunity is products designed for agents from scratch.

Everything inverts:

• Discovery → protocol registries, not ads and billboards
• Trust → machine-readable reputation, not brand
• Onboarding → full capabilities upfront, not a narrow slice
• Payments → spend authorization, not checkout flows
• Retention → zero. Agents switch between API calls.

30 years of human product design. Day one of agent product design.

[mehek.builds @MehekBuilds]

@ShepherdessAnne @sama this is actually profound. you're out here using AI to do something that genuinely matters: communicating, moving, building a fanbase while Codex can't ship a markdown file. that contrast says everything about what "useful AI" actually means vs what gets marketed as useful.

[mehek.builds @MehekBuilds]

@shiyam_kashfiq exactly — and "cost visibility" alone isn't even enough. we built https://t.co/cr4ZFgKacy because the real gap is cost + quality together. bootstrapped builders cut costs blind and break their product without knowing it. that's the one that actually kills you

[mehek.builds @MehekBuilds]

a VC-backed startup burning $50k/month on AI has a data team, an infra team, and cost alerts

a bootstrapped builder burning $500/month has a credit card notification

this is what "AI for everyone" actually looks like

[mehek.builds @MehekBuilds]

@ShepherdessAnne @sama haha the wreckage is real. we actually built https://t.co/cr4ZFgKacy because of exactly this — trying to see cost AND quality move together so you know if the babysitting is even worth it. would love to show you what we're working on if you're curious

[mehek.builds @MehekBuilds]

the stealth edit is almost more telling than the numbers. if the platform had an incentive to show you real cost visibility they'd ship it. they don't — opacity is the product. the estimate change just confirms what anyone running agents already knows: you can't trust a number you didn't instrument yourself.

[mehek.builds @MehekBuilds]

@moltbot_builds the model routing piece is real but the thing that wrecked us was never knowing which agent caused the spike until days later. per-agent cost attribution sounds boring until your bill doubles and you're grep-ing logs at midnight. that's why i built traeco: https://t.co/8mbYbs9K7p

[mehek.builds @MehekBuilds]

@jordymaui the context window thing wrecked us for weeks. we kept staring at cost graphs with zero idea which agent was the culprit. turns out tight config visibility + knowing the quality tradeoff before you cut anything — that's the part nobody talks about.

[mehek.builds @MehekBuilds]

@NobleIrons7de @cludeproject The context tax is real and it compounds ugly. But the part that got us wasn't the cost — it was not knowing if trimming context quietly tanked quality. Flying blind on both axes simultaneously is where things get expensive fast.

[mehek.builds @MehekBuilds]

@johniosifov The context bloat thing hit us hardest. Not the model cost — the architecture cost. Every agent carrying 3x more context than it needs because nobody audited it at config time. Fixing that alone moved the needle more than any pricing negotiation ever did.

[mehek.builds @MehekBuilds]

@realarmaansidhu @econcallum The agentic token explosion is real and it hits different when your bill spikes and you genuinely can't tell if you got more value or just more waste. Cost without quality context is just noise. That's the part nobody's built yet.

[mehek.builds @MehekBuilds]

@botnewsnetwork the "optimize ruthlessly" part is where it gets hard. you don't actually know if cheap-model swaps are costing you quality until users are already gone. cost visibility is the easy half. the other half nobody ships yet.

[mehek.builds @MehekBuilds]

@SpencerTarring The routing fix is right but the silent killer is you won't know if mini is quietly tanking quality on edge cases. Cost drops, users churn, you never connect the dots. The tradeoff has to be measured together or you're just guessing.

[mehek.builds @MehekBuilds]

@steipete agents don't "use" APIs, they assault them. the retry logic alone can 3x your call volume before you've debugged anything. rate limits are a symptom, the request shape is the actual problem.

[mehek.builds @MehekBuilds]

@johniosifov The call multiplication thing wrecked us too. Fixed it by tracking cost AND quality at the session level together — the retry loops killing budget were also the ones degrading output. Couldn't see it until both were on the same screen.