brettcvz

I set up a Hermes agent for our company and connected it to BigQuery, Stripe, Amplitude, etc so that our nontechnical teammates can ask and answer their own data questions via Slack. It works great!

I similarly haven’t found it useful to run a bunch of these agents in developing the product, but it’s a nice interface for people who don’t live in their terminal all day.

I did some research into this - the public budget is actually reasonably detailed. The biggest gap between your analysis and the actual expenditures are the SFUSD faces much higher facilities costs, higher admin cost due to Teacher coaching, and specialized programs for English language learners and special education

Finally, I have no idea where people are getting $28k/year; most schools in CA operate on closer to $14k-$16k per pupil

Also this is an area where first principles analysis is likely to lead you astray - I’d recommend starting with SFUSD’s public budget to understand what their cost structure is.

The big thing you’re missing is special education, and to a lesser extent English Language Learners. School districts are obligated to teach every student, some of whom cost the district dramatically more than they receive from the state.

Your admin costs are also low - you need to account for each teacher being coached and managed, running school operations and front desk, facilities management, finance, IT, etc.

This seems like the judicial branch just voted to give itself substantially more power.

Are there any checks against this? Or can justices just keep granting themselves more powers and invalidating any restraints?

demand decent education

Yes, but with 125k schools in the US and 3.5 million teachers, how do you determine which of them are providing a decent education?

I agree gaming the system is always a temptation with accountability systems.

I’m not sure it would work like how you are proposing, since purposely stunting a student would penalize the school exactly as much as they would “gain” as the student catches back up.

Strong claim! Can you elaborate?

Also, the SBAC test administered in California, Washington, and many others is now adaptive, but still schools are measured based on point-in-time aggregates.

For those interested in this topic, I’d strongly recommend Bryan Caplan’s “The Case Against Education”: https://press.princeton.edu/books/hardcover/9780691174655/th.... He does a much more thorough analysis of the value of a college degree than any other I’ve seen, including factoring in dropout rate, majors, meeting potential spouses, opportunity cost, university prestige, etc.

The summary aligns with common sense: college degrees are typically valuable, especially in in-demand fields and from prestigious universities. But the college dropout rate is around 50%, so many people incur the cost of college with little of the benefit.

If you’re a high schooler considering college, it’s worth it if you are “good at school” and so confident you can finish in four years and are going to a high quality state school or better. Otherwise, you’re better off going directly into the workforce.

Your comment made me think as well! I suspect we might be caught in a definitional discussion of the word “feedback” - feedback is a subset of “words someone says to me”.

My working definition of feedback is “things I tell you about something you did with the intent to help you do better next time,” which does warrant the “feedback is a gift” mentality, even if the feedback is poorly delivered, misguided, or wrong. The opposite of love is not hate, it’s apathy.

As other have mentioned, sometimes people say things to you that are not feedback, and are primarily them seeking status, expressing hurt, etc. a caution though that “this person is just having a bad day” is an easy, universal response for dismissing valid but critical feedback.

I understand the phrase “feedback is a gift” differently than the author: people are not obligated to give you feedback. If someone has taken the time to give you feedback, this is a generous act, regardless of whether you agree with the feedback or find it helpful.

Like many others, I think this is super cool, and something that I think has been missing in the ecosystem for a long time.

I feel like the thing that would really make this over-the-top powerful is deep integration with a component library, either a company's in-house one or an open-source library. It would allow for super fast UI prototyping that would also serve as scaffolding for the full-featured product.

Likewise re. Jonathan - we sold filepicker.io (now Filestack) to Jonathan and Xenon Ventures in 2014 and it was a very positive experience. For companies with solid revenues but not venture-scale growth looking for a clean exit, I would highly recommend reaching out to Jonathan and team. Happy to make an introduction if helpful.

As PG recently shared, when an investor puts money into a company it is a calculated bet that the company is actually worth _more_ than the valuation they are investing at. No one invests $1 for a 10% chance of making $10. So if the valuation goes up, it basically eats into an investors expected “profits”.

I write at http://brettcvz.com - I primarily write about “teachable points of view” I’ve learned about product management and startups, along with an occasional electronics project.

I link to a few helpful sources in the “references” part of the project readme: https://github.com/brettcvz/upnext/blob/master/README.md#ref...

Hope that helps!

Agree, the waveshare modules are great. They’re also (surprisingly) available via Amazon Prime, which is nice when you’re excited about a project idea and don’t want to wait 2-3 weeks for a shipment from China.

From my experience with this project, all ePaper displays _can_ do partial refresh, but some may have the drivers (LUT) included, others you might have to find some community-sources ones or design it yourself. In the post I link to my drivers for partial updates for the 4.2”, and in the footnotes there is a YouTube video going into lots of detail on how the drivers work if you wanted to write one or update it for a different display.

Thank you! I’m not a designer, so the advice of my designer friends was “pick exactly 2 fonts, use lots of white space, and left align edges.”

The code for the site is open source, feel free to borrow/steal.

This was a labor of love, so it’s definitely not that I don’t have time - more that no project of mine has ever been this popular, so normally no one asks for the tutorial!

As others have commented, this project is mostly software and would be a great introduction to building physical electronic gadgets.

How about this - I think there’s enough context in the blog post and source code to recreate it, but if you start out and run into any issues, send me an email and I’ll get you unstuck. In exchange, I’d ask that you take detailed notes of your process and post a blog post/github readme of how you got it running for others to follow, and I’ll link to it.

You are correct - because I have a clock on the screen, UpNext does a full screen refresh about every 10 minutes. I thought this would bother me so did some explorations of a UI without the clock, but in practice I don’t notice the flashing, and I like having the clock.

I can provide some context here - from my experimentation, after 10 partial updates the contrast is notably worse, but doing a full screen refresh resets things sufficiently that the screen looks sharp again.

To reduce degradations further, whenever UpNext boots it does a cycle of 10 full screen refreshes - overkill certainly, but I think of this as the “cleaning cycle.” If the screen ever starts to fade, I power cycle it, and it’s fine for a long time.

I’ve been running it continuously like this for 2 years now, and while the screen is a bit less sharp than it was originally, it’s still in good shape - the photos in the article are from this week.

It’s this one: https://www.amazon.com/gp/product/B075JLP2LP/

I suspect that the price of the components put this out of the range that most people would pay for it, but I appreciate the sentiment.

While most of the IP at this point is open-source, if someone is excited about figuring out how to turn this into a kickstarter/actual sellable hardware product, I'd be willing to collaborate.

Very kind of you! I'm happy to help you set one up if you're willing to buy the components. The wiring is very simple, and while the code is open source, I'd also be happy to send you the SD card image if you'd like.

It’s a neat idea! We have an “internal products” team at Clever that is tasked with building tooling for our support team (and all our other teams too!) that makes them more efficient. They are definitely worth the investment, and I’d encourage the approach to other companies.

Re. Bugs - I actually really like the “top 2 things” approach for exactly this reason. It’s very illuminating when you as the PM are like “look at all these exciting things we have on the roadmap!” and your support counterpart responds with “my #1 request is that you fix this bug.” This is why I think that the customer facing teams can sometimes be better at doing this prioritization - even with lots of data about the size of a problem and anecdotes the pain, it’s still easy as a PM to not give it the appropriate attention, whereas the support team that lives with the pain will fight for it.

Do you know anywhere that does this really well in practice? My concern is that it would end up similar to the internal feedback problem where people feel like they submit requests into a queue that never gets worked on. I feel like I know how to solve that problem when there’s a small number of internal stakeholders I can go over and talk to, I’m not sure how one does it at scale with users

Totally agree with the “yes and” approach of firehose plus top two - I haven’t had a chance to update the post yet, but responded to similar feedback here: https://news.ycombinator.com/item?id=21971011

Totally agree! I use this technique as a complimentary layer to “send all possible feedback my way”. Tactically, we achieve this through having a handful of different “feedback-X” channels for different audiences that people dump feedback into. But over time that can start to feel like a black hole that isn’t listened to, so I like adding this layer on top. Good point though, I can add a clarification to the post.

Okay I think we’re saying pretty similar things then!

The “top two” technique works particularly well for addressing “performance” type features for products already in market. Most customer requests, and most of the things that my stakeholders come to me with, are around ways the current product could better address the existing problem it’s trying to solve, rather than finding whole new customer pain points.