dchuk

In reading the docs, it doesn't look like the MCP server supports the Ask Sourcebot capability. Is that correct or am I missing something in the docs? Is that planned to be added?

I’ll admit this looks comprehensive, but man oh man does this seem complicated and over doing it

Your build guide link is a 404

That’s the exact point?

Well done site. One important nit pick: never use charts that don’t start from 0 on the y axis. I was looking at a stock that had a yoy growth rate reduction of 6% (from 39 to 33 for each respective yoy period), and the chart showed an aggressive down to the right trend line because the y axis started at 33% instead of 0%.

Charts like that show more detail sure, but everyone freaks out in reaction to them. Always zero out your graphs.

Who on earth knows what 14g of alcohol is in terms of visiting a bar?

Terrible time to go to Disney, late March early April is spring break for basically the whole country

While I don’t disagree that Microsoft isn’t necessarily the most innovative, Apple is absolutely the most successful example of “first mover advantage” not actually being an advantage, and second movers can often be the actual winners. Literally every product they have launched in the last few decades has had the category already defined, they just came in with so much more polish and elegance (and sharp marketing) that they just repeatedly obliterate and take over the category as if they invented it.

I’ll never forget when they dedicated a minute or two in a keynote a few years ago to how they improved the volume indicator visual overlay in iOS to be less obtrusive like they talked to god himself to figure this out, when Android had that style for years…it was brilliant marketing.

I say all of this as a die hard Apple guy.

I’ve been tinkering with Marimo, it’s pretty sweet (and you can use cursor or other AI IDEs pretty easily with it).

On running notebooks as scripts: I can’t find in the docs what happens if you have plotting and other notebook oriented code? Like I’m using pygwalker to explore data through transformation steps, and end with saving to csv. If I just run the notebook as a script, is all of the plotting automatically skipped?

I’ll never understand how any UI projects don’t include an actual screenshot of their project as the first thing on their landing page. It seems so obvious.

Are you chatting with it directly or using it with a coding tool or IDE?

I can’t pronounce two of the four words in that domain, that’s impressive.

You do know sometimes business people talk with lawyers about business topics right? There are business conversations sensitive enough to not be recorded and transcribed by some cloud service that could be breached.

Being able to have windows of the same app assigned to different spaces would be the big unlock for me to able to use your app. Not only do I do things like separate work and personal Chrome profiles (with a window open for each), frequently I'll have a Chrome window open for a specific project with tabs just for that project, so I can switch fully into that context when I want to work on something. I also frequently have sensitive tabs open that I group into a window so I don't risk oversharing something when on a Zoom call for example as I flip through my tabs.

Looks amazing, but the limitations section seems concerning given the nature of the tool:

“Limitations Deep research unlocks significant new capabilities, but it’s still early and has limitations. It can sometimes hallucinate facts in responses or make incorrect inferences, though at a notably lower rate than existing ChatGPT models, according to internal evaluations. It may struggle with distinguishing authoritative information from rumors, and currently shows weakness in confidence calibration, often failing to convey uncertainty accurately. At launch, there may be minor formatting errors in reports and citations, and tasks may take longer to kick off. We expect all these issues to quickly improve with more usage and time.”

Given the point of the agent is to replace hours of manual research and factual distillation/summarizing, the risk of hallucinations and confident falsehoods in the data sort of undermines the whole shebang. If you want rock solid data but can’t afford to spend the time getting it, seems like a terrible trade off to save time but increase risk of bad data.

What I’d love to see is a containerized android that can be fired up on a Mac (using docker desktop or orbstack or whatever) that I can modify the docker image of to have rooted man in the middle proxy already setup, making it much easier to drop an Android app onto to observe the network traffic and api calls.

“Simplified desktop”…go to their website, literally no pictures of the desktop interface (at least on mobile).

Everyone who makes software, always follow this formula on your homepage: clearly state what problem exists without your software, show your software, clearly state what problems go away with your software, then describe its actual features

I travel extensively for work. Where are you guys staying that doesn’t have a basic desk and chair with outlets and stuff ready to go? Even the most basic Marriott property has a totally comfortable workspace in every room

I have zero need for a portable raspberry pi cluster but damn is this cool anyway.

Also, I learned about this device from this post and immediately bought one for my existing home server remote access: https://jetkvm.com/

Here’s a business idea for someone: find the best open source/self hosted business and start up tools (such as a chat app, a project management tool, a wiki, etc) and offer startups a managed hosting service of those selected apps. Toss up a cheap but decent vps per client, run some open source monitoring tools for yourself, price as a flat rate per app under management (vs scaling rate of a per user seat in most SaaS products).

Seems like a pretty damn easy lifestyle business that many early startups would find valuable because it saves them money and the headache of doing it themselves.

This looks awesome.

What I would love to see either as something leveraging this, or built in to this, is if you prompt stagehand to extract data from a page, it also returns the xpath elements you'd use to re-scrape the page without having to use an LLM to do that second scraping.

So basically, you can scrape new pages never before seen with the non-deterministic LLM tool, and then when you need to rescrape the page again to update content for example, you can use the cheaper old-school scraping method.

Not sure how brittle this would be both going from LLM version to xcode version reliably, or how to fallback to the LLM version if your xcode script fails, but overall conceptually, being able to scrape using the smart tools but then building up basically a library of dumb scraping scripts over time would be killer.

I’ve had a similar idea over the years. You should consider exploring whether competitive companies could be customers.

As a competitor, getting alerts about roles another company is hiring for can be very interesting. Combine it with trends of postings over time…

While I agree that just dumping the puzzle instructions into an llm and getting the right answer at least doesn’t align with the spirit of Advent of Code (though it does show how crazy good LLMs are getting), I’m using this as an opportunity to try out the new Windsurf AI IDE and am driving the AI code editing in it to help me write the code to solve each puzzle (I’m using Ruby).

My goal is to develop the muscle memory for the tool so I can write code quickly. I’m still generally thinking through the puzzles, but being able to just write out plain English logic, get code generated, ask for it to be well documented, quickly refactor things to be generally reusable, etc, is just fantastic and how all software development should be done in this day and age frankly. Such an accelerator to problem solving.

At least on iOS, Reader mode in safari fixes it right up

Appreciate the quick iteration!

Am I missing something or does this just completely not do anything on iOS/mobile? I see in the json editor a single object with two keys, and then no visual tree rendering whatsoever

I’ve reverse engineered a few industry conference apps to more easily get the list of attendees (some conferences will literally only give you a pdf of scanned paper lists of contact info for attendees, which is insane, especially if you are paying to have a booth there). I’ve either decompiled the Android app, or ran mitmproxy on the device, or both, to figure it out. Does anyone have a recommendation for having a pre-rooted Android simulator with the tools you need installed and ready to go to make this a quicker process? I’d love to just drag and drop an apk into a simulator and start inspecting vs having to use a real device and all that jazz.

Interesting potential use case for AI: automatically rewrite the code in a book like this for python 3 vs 2. I know that’s not trivial, but seems like a very trainable use case

You desperately need to show your product on your landing page, and have headings that convey the individual use cases and value props of it for the reader.

Simple general rule of digital marketing: No one reads website copy. They look at pictures and scan headings.

Another simple rule: I inherently don’t care about your product if I’m reading your site for the first time, so don’t spend time describing your capabilities. Simply describe how my life will be better if I use your product, and show me the product doing those things.

Less words more pictures. Less words more value.

(Sorry for being blunt, just trying to help with your conversions)

Look into this: https://www.anthropic.com/news/contextual-retrieval

And this: https://microsoft.github.io/graphrag/