phsource

Great call on this! Let me see if I can use a zero-data-retention version of Google Gemini too but if not I can probably add a model pickers

Hey everyone! My name is Peter, and I'm one of the co-founders of Wanderlog (YC W19). After reading lots of articles on redaction failures, including most recently the one about Google's data center water/power usage [1] (which admittedly is not that big of a leak), I wanted to build a better tool that makes redaction both not tedious and effective.

RedactPDF.ai uses Gemini 3.8, which is very good at recognizing data to be redacted and drawing the correct bounding boxes for it. It automatically adds redaction boxes for the data you might find sensitive. My goal here is eventually to make it easier to share data and document templates that can help people avoid the tedium of filling out forms and applications.

I'd love to hear your thoughts: the code is open source; I tried to make it fully local-only, but unfortunately, it's still a little difficult right now, given that other models are not quite as good as Gemini 3.8. Try it with your own documents and let me know if you run into any issues. I'd be happy to tweak the prompt or take pull requests to improve the process!

[1] https://news.ycombinator.com/item?id=49957068

Simon Willison's post about this gives a good context on why exactly this is happening. While it doesn't mention this in the Artificial Analysis page, this is likely with Max reasoning, which has extremely long reasoning traces:

https://simonwillison.net/2026/Aug/16/qwen-38-27b/

It seems like the token usage is 2.3x GPT Luna Max and almost 2x Kimi K3!

https://imgur.com/a/dDSyhr2

I'm curious if they can make up for this with insanely high tokens-per-second especially when served from hosted providers, though, given how tiny it is (37B!)

Personally, with our company on Cursor, I can see why model makers are not the best people to go all the way down the stack. Using the right model for the situation will continue to be important, and model makers, by design, do not want to give you the choice to run different models.

Right now, we use:

- Kimi K2.5 for easy fixes, asking about the code, various agentic commands (e.g., summarizing Loom videos for Slack messages)

- Opus 4.8, Sonnet, or Kimi for planning (we find GPT-5.5 to have too terse outputs for plans)

- Kimi K2.5, Composer 2.5, GPT-5.4 mini, etc. for faster implementation (i.e. we don't have to wait around for the slower tokens-per-second generation on Sonnet, etc.)

If we had to only use Opus, Sonnet, and Haiku, I'd definitely be looking to switch harnesses

This is so neat! I really think AI turbocharges this kind of personal project way more than it speeds up programming for work:

> I was curious about the possibility of doing this myself, and I asked ChatGPT. Not surprisingly, it knew a lot of the various tapes, file formats, sizes, processing, storage, and after it asked some clarifying questions, it was quite optimistic about me being able to do this myself

Between this, it seems like it helped with so many different parts of the process:

1. Asking for how to do technical things, like transfer video from these old VHS to a newer computer.

2. Writing code for the web portal to host the videos.

3. Writing VLC plugins to help with data entry.

4. Transcribe audio into text.

Similarly, a coworker recently made a website that imitates what Alpha School does to incentivize his own kids to finish their homework all in the span of a weekend, and it's cool to think of the kinds of projects that less or minimally technical people can do with the help of ChatGPT to guide them.

Of course, the debugging techniques and the debugging and problem-solving techniques that you get from being a professional programmer helps a lot with taking what LLMs give you with a grain of salt, and knowing what they're good at and what they're not. But it is a superpower for sure.

This is pretty impressive results given that this is not from one of the major AI labs. Congrats: https://blog.withmeka.com/meka-achieves-state-of-the-art-per...

Out of curiosity, what do you think contributed to this working better than even OpenAI agent or some of the other tools out there?

I'm not that familiar with how OpenAI and other agents like Browser Use currently work, but is this, in your opinion, the most important factor?

> An infrastructure provider that exposes OS-level controls, not just a browser layer with Playwright screenshots. This is important for performance as a number of common web elements are rendered at the system level, invisible to the browser page

Hey figmert -- this is Peter, one of the co-founders of Wanderlog. I'm actually on a trip to Italy right now and definitely feel your pain with some performance issues, and we've been working hard to improve this.

If you haven't tried the app in the last few months, can you try it again and let me know what parts are feeling slow for you by emailing me directly at [email protected]? I'd love to take a closer look, and especially if you've got specifics with screenshots/videos, I can try to fix some of these myself too.

We've had to go through this process for the app I have, and it definitely was cumbersome and makes the process a huge pain. Fortunately, after a while Google often lets you switch to a Tier 1 assessment, which involves using various tools to analyze your code and make improvements without shelling out a ton of money.

At the same time, Google is in a tough spot here. The files and documents in your Google Drive (or Gmail) are incredibly sensitive. One possible solution is using the https://www.googleapis.com/auth/drive.file OAuth scope, which only lets you access files a user has explicitly shared with the app. I'm curious if iA Writer has limitations that makes this a bad user experience, but from a user security point of view, I can see why I want the apps that get to see my whole Google Drive audited too.

[1] https://developers.google.com/drive/api/guides/api-specific-...

This is a bit of an unusual choice, but I recently was trying to use Webflow, and its videos are both highly informative and funny! Engineers often thumb their noses at videos, but they really show how to do it well:

https://university.webflow.com/videos

Cool stuff! How does this compare with Fastchat, which seems like another open source project that helps run LLM models?

At a glance, it seems like it's going for lots of similar goals (run LLMs with interoperable APIs):

https://github.com/lm-sys/FastChat

Just to be clear, the "capital" here is not really money, but instead, machines and equipment. Think a car company (e.g., Tesla) buying a used car factory (i.e. the NUMMI plant [1]) or a biotech company buying equipment cast off from Genentech or Pfizer.

> To document the interaction between firm and machine age, we lean on 1.56 million transactions covering 70,000 models of machines. Across a wide range of industries and equipment types, young firms acquire older capital, whereas older firms are more likely to buy new capital

This seems to line up with what I've heard from friends in biotech and mechanical companies. One friend lamented that "at Apple, we had the best of everything: suppliers, equipment, machinery. Here, we make do with what we've got and find creative ways to work with them." Similarly, on the YC forums, we often see used lab equipment listed for sale, and there are whole industries around this.

Pretty neat! Definitely a part of the startup or industry network effect I haven't thought of before. I wonder what other examples of startups working with super-janky second-hand machinery are out there.

[1] https://en.wikipedia.org/wiki/NUMMI

I'm sad to see all the hate this article's getting in the comments, but also can't say I expected anything else.

Most of the hate I'm seeing here is focused around work-life balance, but even the author acknowledges that "I don’t think people should cry or feel like impostors or skip their vacations regularly." There's so much more here! Kudos to Brie too for really putting herself out there like this.

Taking apart what the article mentions, the fulfillment she felt at work came from:

- A sense of shared mission: of the mission statement, "it was a little abstract, but we believed in it enough to recite it with pride"

- Being pushed to do better: "My work was meticulously but warmly critiqued by my peers and leaders alike, and my work got better and better because of it"

- A community and culture: "It felt like magic, but there was deep thought, care, and intention behind everything. I had a tingly feeling that I was part of an organization that had cracked something about creating a great culture"

Forget work -- think about a side project, a hobby, a sport -- anything that you've applied yourself to. Does it feel good to hold yourself to a high standard, in the company of other peers who are into the same things?

That sounds like something we can all get behind!

Stripe may not have been perfect, but let's not throw the baby out with the bathwater here.

Life insurance is actually surprisingly common a perk; I remember even Stripe had it! I guess it's nice to know that your employees' family will be well provided for regardless of their original financial situation.

Maybe this is a healthy dose of paternalism? I feel like there would be more people who regret not buying life insurance in this case than people who regretted the minimal $250 outlay

> For example, the platform draws on historical support tickets and even Reddit activity to anticipate future volumes of customer requests (by queue, channel, and site).

Wow, do companies now use Reddit for support? I'm curious how exactly Reddit comes into the customer support volume estimation process

Yeah, that's pretty much it -- usually nobody stress-tested this before they just sent it out in a rush to employee #1, who's probably a referral anyways

> If the employer is a good one, they want the contract to be fair, and if the contract is unfair it's probably by accident.

This definitely sounds about right -- I'm pretty sure most places I worked and most other Y Combinator companies in my batch just took the standard forms from somewhere (e.g., the Orrick forms library) and used them, without thinking too much about the terms. [1]

If there's something in the terms that actually seem objectionable, it's likely the hiring manager, recruiter, and other people in your recruiting process have never thought about it either. By pointing it out, you're often educating them too on what the offer letter says, and providing valuable feedback for the recruiting process.

[1] https://www.orrick.com/en/Total-Access/Tool-Kit/Start-Up-For...

This 100%. In travel, we see Google constantly tweaking its algorithms, and compared to Bing, Google surfaces a ton more small, well-written travel blogs [1]

Not only that, Paul and Michael have seen plenty of startups, and at least in recent memory, the number of vertical search and consumer startups that Y Combinator has funded hasn't been that high

As a consumer startup, I know this issue firsthand. Paul and Michael assume that if you build a better product, they will come! That's simply not true these days.

Instead, you need to:

- Build a better product

- Option 1: Figure out a channel with enough growth on an existing platform. This likely means you're doing SEO for your new search engine

- Option 2: Get your customer lifetime value high enough so you can pay for ads. This is tough, since it's a bit of a chicken and the egg problem since most search engines are monetized with ads

As the founder of Wanderlog (YC W19; https://wanderlog.com), a consumer vacation planning app [1], I definitely remember the idealistic days when I thought the best consumer product on its own would win! But growth doesn't just come, and the same can be said of vertical-specific search engines.

[1] Try searching "[your city] itinerary" on Google vs. Bing: it's much more likely you'll find a small blog rather than Lonely Planet or the local travel bureau as the top result

I think it's interesting that the author mentions the Playstation Portable as an example: it seems like the reason it doesn't work actually _isn't_ a fault of the device manufacturer:

> You might think that a 15-year-old gaming console wouldn't even be operating, but Sony's build quality is such that, with the exception of a very tired lithium-Ion battery, the unit is in perfect condition. It runs but can't connect to modern Wi-Fi without an update, which it can't access without an update to its firmware (a classic catch-22).

In this case, developers seem to actually have provided updates, but it just doesn't work! Reminds me of when I booted up an old laptop to find that the old root SSL certificates don't work for sites today and everything's HTTPS now ¯\_(ツ)_/¯

It really points to an issue with net-connected tech (and yes, the PSP actually benefits from being online unlike certain ridiculous IOT devices): really, even if the manufacturer designs with long term support in mind, the rest of the world doesn't. So I don't know quite how tractable a problem this really is in the Internet age; the solution might be just better electronics recycling rather than indefinite updates

This is on our roadmap! A bunch of folks have asked for it, and there are good APIs

Oh boy -- I'm the cofounder of Wanderlog (https://wanderlog.com), a YC W19 startup exactly focused on planning travel. And I've really read all these articles

This originally was big on Hacker News in 2014: https://news.ycombinator.com/item?id=8419658

This is really just the latest in a long list of articles about why travel planning startups aren't worth pursuing:

- https://www.phocuswire.com/Why-you-should-never-consider-a-t...

- https://paansm.medium.com/the-top-5-reasons-your-travel-star...

But we're still at it and have grown a ton over the past year!

- On frequency: I think we're underestimating folks' ability to remember here! If the product actually is good, people will remember it.

- On being a real pain point: It's true that people have many "good enough" solutions to travel planning, but travel planning tools have failed because the past products really just haven’t been better by enough than using Google Sheets -- they assume that people don't like planning, or that they plan a certain way, and try to save time and automate it. In fact, people _love_ planning and sharing their trips! The tool just needs to work around the traveler, rather than dictate

Still, it's instructive that none of the links in the 2014 post's comments work. Will we prove them wrong? We can check back in 3 years when this article comes back to the front page again xD

This might as well be called the Age of the Blog Post, which seems obvious today in our age of Substack, Medium, and Wordpress, but back in 2004, definitely wasn't quite as clear!

> The Internet is changing that. Anyone can publish an essay on the Web, and it gets judged, as any writing should, by what it says, not who wrote it. Who are you to write about x? You are whatever you wrote.

That being said, I think the above has contributed a bunch to sensationalism. Unbundling of writing from magazines and newspapers into little articles incentivizes writing individual articles that stand out -- clickbait, shockers, and half-truths.

Good writing and crisp thinking definitely still rises to the top, but so does controversy and stirring up emotions. So yes, essays have won, but it's less clear the kind of essay that Paul mentions ("a few topics you've thought about a lot, and some ability to ferret out the unexpected") are the ones that have won