General
p/generalShare and discuss tech, products, business, startups, or product recommendations
trending
•

4mo ago

How do you actually find roles that never get publicly posted?

I've spent the past few weeks digging into how people land jobs that never make it to a job board trying to separate the tactics that actually work from the recycled "80% of jobs are hidden" stat that gets repeated a lot without much evidence behind it.

For those who've actually done it: what worked? A referral, a cold email straight to the hiring manager, a tip from someone inside? Trying to figure out which of these still hold up in 2026 versus outdated advice.

•

4mo ago

Designing a well rounded API

I got recently an request to build an API from a user and I wanted to ask if people here have experience in designing one. Earlier we've only considered building a pull API, but this user requested us to build specifically a push API.
This got me to thinking that I don't know nearly enough about how to design a well rounded API. What I'm wondering about is:

  • While I want to give our customers maximum privacy while using our API, I also want to have the ability to detect issues and potential attacks in real time. Are there best practices regarding analytics for the API usage?

  • Usage based pricing is all the rage now which makes sense with AI agents and all. But I'm thinking about a flat monthly fee with some (upgradeable) limits, particularly considering the push model. Would really appreciate if anyone could share their experiences on how to discover the right pricing model for an API?

  • I've also seen a few API providers making a Model Context Protocols to deliver their data once the user has subscribed. Any experiences with making them (and providing data through them as well)?

Really appreciate any help and insights!

The best onboarding test is whether users can predict the next screen

Polished onboarding can still be confusing if each step feels like a surprise. After the first action, ask a new user what they think will happen next. Their prediction exposes whether the product's mental model is clear before completion rates do.

If five people finish but all five expected a different result, the funnel is technically working and the product is still teaching the wrong thing.

Where in your onboarding do users most often predict the wrong next step?

A benchmark without a failure taxonomy is just a scoreboard

A single score tells you who won. It rarely tells you what broke. For model or product evaluation, I want failures grouped before I trust the ranking: factual error, stale source, bad tool choice, latency collapse, unsafe action, or a task the system should have refused.

That taxonomy changes the next engineering decision. A two-point gain driven by easier cases can hide a worse regression in the failure mode users actually care about.

What failure category has changed how you read a benchmark?

A model architecture claim should name the active compute

Parameter count alone has become a bad shortcut for explaining conditional models. If a system uses experts or sparse memory, I want four numbers: total parameters, parameters active per token, routing or retrieval overhead, and measured throughput on named hardware.

Without those, a large number can describe storage capacity while readers assume every parameter contributes to every token. That muddles both cost and capability.

Which number do you wish model announcements reported by default?

The first user interview question should be about the last attempt

Ask someone whether they want your idea and you will get a polite forecast. Ask them to walk through the last time they tried to solve the problem and you get evidence: the trigger, the workaround, the cost, where it failed, and whether the problem was painful enough to revisit.

Specific past behavior beats imaginary future enthusiasm. It also makes the awkward answer useful when they did nothing at all.

What question reliably pulls your interviews out of opinion mode?

Your launch retrospective needs the people who left

Launch retrospectives over-sample the people who stayed long enough to talk. The most flattering feedback is also the easiest to collect.

I want the quiet exits in the room: people who clicked but did not start, started but never reached value, or used the product once and vanished. Even a tiny exit survey paired with the last completed step can reveal more than another call with a power user.

What is the earliest silent-exit signal you review after launch?

A study streak should survive a bad week

A streak that resets after one missed day teaches the wrong lesson: consistency is fragile, so a setback might as well become a week off. Real learning has exams, illness, work shifts, and plain bad Tuesdays.

I prefer a recovery rule over a perfect chain. Miss a session, then return with the smallest meaningful action: retrieve one idea, solve one problem, or name the next gap. The habit worth reinforcing is coming back.

What would you measure instead of consecutive days?

•

4mo ago

Do you ever work on your startup all day and still feel like you did nothing?

I've had a lot of days like this recently.

I can spend the entire day fixing small product details, coding, replying to people, reviewing designs, making designs, posting on social media, writing copy, planning features, discussing decisions with my co-founders, testing something, and solving random problems that appeared five minutes earlier.

•

4mo ago

At what point do you hire someone just to keep track of things, versus fix the process instead?

I keep meeting people running one or two service based businesses who reach the same fork in the road. Hire an ops person to catch what is slipping, or find a better way to catch it themselves first. I went the second route out of necessity, cash was tight and a hire felt too early. Curious where others draw that line, is it revenue, headcount, or just how much you personally cannot hold in your head anymore?

First
Previous
•••
554555556
•••
Next