randomdrake

Why wouldn’t they? After reading Careless People I’m not surprised anymore when Meta sinks to lows like this.

Why? They’ve done it before. The punishments did not outweigh the advantages. Don’t forget Myanmar, or the United States, or countless other countries where meddling has occurred in relation to public change or elections.

I deleted all Meta products from my phone and am waiting on my account export before removing Meta from my life completely, after nearly 20 years.

They could’ve become something very different. I celebrated Zuckerberg ringing the opening bell thinking, “finally, a geek got his day.” Little did we know.

They chose this.

After a bit of research, it looks like BOINC[0] and AI Horde[1] are along the lines of what I was building. It appears NVIDIA PAIR[2] has been working on distributed inference for local networks as well.

[0] - https://boinc.berkeley.edu/

[1] - https://aihorde.net/

[2] - https://www.nvidia.com/en-us/ai-on-rtx/personal-ai-router/

Yes! I used to love watching the visualizations for this. You’ve got the right idea. If it truly is AI processing to help us eventually create this virtual biology, why not our idle machines with models running on them?

You’re right, of course. I never made it out of PoC. But I also implemented something with locally running models to do the same thing. Request comes in, gets forwarded to a machine not currently busy, it serves.

It would be great if we could use the credits we pay for and don’t use on our subsidized subscriptions as well. But, yes, a matter of “when” not “if” for your account getting shut down.

I’ve been toying with the idea of a SETI@Home successor that uses leftover subscription credits from the (currently) subsidized AI providers to help contribute to a collective goal like this.

I wonder if there will be any opportunities to combine resources collectively like that again in the future.

I built a small prototype that would let people in less-developed countries get AI responses from it for free. But it’d be super neat to work together and put our laptops and subscriptions to work on something like this.

How can we determine whether it’s negligence, or advertising? Do we give the benefit of the doubt? Apply Hanlon’s razor? What would we lose, in terms of capabilities, if the underlying behavior wasn’t allowed to accidentally happen? Would we have worse models than we do? There are so many questions. I’m afraid that, until publicly disclosed discovery occurs in one of these cases, we’re just guessing for so many important conclusions.

I wonder if this extends to training models on new content as well. Are we creating a cyclical information-consumption and training situation in which models being trained are more likely to pick up on and reference content created by themselves or by other LLMs than by other humans?

  Location: Tigard, OR (Portland Metro)
  Remote: Yes
  Willing to relocate: No
  Technologies: TypeScript, Python, Go, PHP, AWS, Kubernetes, React, Angular, Postgres, MySQL, Claude / MCP infrastructure
  Résumé/CV: https://davidbyrondrake.com/david-drake-resume.pdf
  Email: [email protected]

Engineering leader with nearly 20 years building teams that ship. Looking for Engineering Manager, Director, or VP roles.

Most recent: DevOps and Engineering Manager at Counterpart (insurance), where I led two full-stack teams plus DevOps and built an MCP infrastructure connecting GitHub, Linear, Salesforce, Snowflake, and Slack across the full insurance lifecycle. Agile success rate peaked at 89% against an 80% industry benchmark. A git-tracked Salesforce deployment pipeline I built on GitHub Actions saved roughly $20k a year and made deploys boring again.

Before that: Director of Engineering at Flourish Software (up to 15 engineers, 60% improvement in on-time delivery, 50% reduction in support load), Senior Full Stack at Expensify (moved millions of dollars a day, six-stack infrastructure), CEO of RedDep Labs (17 people, $8.6M valuation), CTO at iCracked (Y Combinator, 400+ technicians, dispatching system).

How I lead: I hire people who care about each other's success as much as their own, set clear commitments, and then trust the team to do the work. I trained as an actor and was a special education teacher before I wrote code, which is how I keep technical conversations legible to non-engineers and how I'm comfortable in any room.

Best fit: a team that values shipping over performance, has real users, and wants a leader technical enough to read the code and grounded enough to keep the team steady when things get hard.

I got a paper route just to get a hold of the dx2.

It was a life-changing machine.

Ordered, I believe, from the depths of a Computer Shopper magazine.

When I was in college, we were required to take a business class (Business 101) that mandated a finished business plan as part of the project.

It had to be long, in-depth, and include everything you mentioned.

I was incredibly surprised when I entered the tech and startup workforce that these were generally absent.

I had misunderstood the class and instructor and thought that you couldn't even start a business without one.

Then, when I started raising money for my own venture, I thought for sure a complete business plan was a prerequisite.

Nope. A few graphs, preferably hockey-shaped, and a good story were all that was necessary.

My venture failed, of course. But if I were to do it again, I would do myself the favor of having a complete plan. It would definitely save a lot of headaches and guessing in the moment.

Thank you for your post. Very informative. Why is it too early for AI? It’s clearly an emergent cultural evolutionary byproduct that’s been many years in the making and quite mature. Perhaps your own bias is limiting you to imagine what AI is truly capable of?

Wrote and released a daily reader on sobriety and stillness called I Will Not Drink With You Today.

There’s a companion website: https://iwillnotdrinkwithyoutoday.com

I wrote the book in markdown, stuck it in a SQLite DB and wrote a parser to put all the data in static JSON so it loads very fast.

I also created a new personal homepage to update my presence on the web as a published author and experienced leader and technologist: https://davidbyrondrake.com

Book was released less than a month ago—growing it organically like a startup has been fascinating in terms of marketing, sharing, building, and measuring success.

Have been utilizing my acting skills again with readings from the book on my Instagram and TikTok.

Having a really good time with it!

Defending focus is way harder than adding features.

When you're building, adding yet another feature can sometimes shave off all the edges that made you successful in the first place.

Same with messaging. The more you try to sound universal, the less anyone hears you.

Strong opinions that are honestly held and communicated are such great signs of respect. It's refreshing to see: "This is who we are. If it's not for you, that's okay."

Good piece.

This is a well-done library that’s fun to work with. I put together a proof of concept org chart generator[0] with it a while back when working on another project. Very easy to use and well-documented.

[0] - https://github.com/randomdrake/react-flow-org-chart

Betty White holds such a highly regarded “Hollywood Star” place for me. It was fascinating to see her brought to life through her very ordinary belongings. Fun read.

The original research[0] is available for free online.

It goes into more detail about how they established the genetic signatures and has a lot more images as well.

Epigenetic signatures of intergenerational exposure to violence in three generations of Syrian refugees

[0] - https://www.nature.com/articles/s41598-025-89818-z

Worked here. Changed my mind on SQLite. Hipp worked alongside us. SQLite can be crazy performant:

https://use.expensify.com/blog/scaling-sqlite-to-4m-qps-on-a...

Nice to see the library brought back in some kind of way for tech companies.

15 years ago, most tech companies had a library of sorts because it was required. Sure, there were manuals online, but they didn't offer the depth available in books at the time.

Stack Overflow either didn't exist or was in its infancy, so your options were to either find an O'Reilly book on the subject[0], hop in a Freenode channel on IRC and talk to folks who were in the weeds (or even the maintainers themselves), or you had to use mailing lists.

[0] - okay, not all of the books were O'Reilly but a huge number of them around that time period were

The arts. The further and further you go in instruction, the more it becomes about the little differences and quality. Practice always helps, but quality definitely taught and learned by many as well.

Fantastic claims require fantastic evidence.

Please be careful with things like this before someone gets hurt or worse.

I have friends in the medical profession who have tested GPT4 and it’s good, but not quite good enough for them.

I wouldn’t touch this with a 20 mile stick; something that claims actual medical advice like this.

I don’t know about the method of application, but the restaurant industry, in particular, has almost always operated on hiring bodies that actually walk in the door versus applications from someone online. Especially true for entry-level stuff in my experience (10 years in the service industry).

> Retail and restaurant companies won't ever respond to my applications, despite constantly spamming the same jobs month after month.

I don't know of any place in the U.S. that doesn't have wanted signs in the windows for restaurant work. Between fast food, back of house, or front of house, there really are a ton of options both in downtown areas and suburbs from my experience.

Have you actually gone to one of these locations, physically, and applied? Did you follow up instead of just dropping an application by and hoping for the best?

Examining the source, you can use coordinates instead of address when calling get_aoi.

https://github.com/chrieke/prettymapp/blob/main/prettymapp/g...

Sounded like a neat project but seemed light on documentation.

Had to go back to 2015 in the Internet Archive[0] to actually find a copy of the website mentioned.

The site is in Russian but Google Translate seems to do an okay job of bringing it into English[1].

[0] - https://web.archive.org/web/20150223174714/http://www.hydron...

[1] - https://web-archive-org.translate.goog/web/20150223174714/ht...

Totally fair. I guess I was mostly considering the more typical client-server consumption, RPC-type uses and wasn’t really thinking just how much we use it for basic data interchange and serialization. Thanks.

How far from WSDLs and SOAP we’ve come.

I applaud this effort. I hate that we’re back to where this feels necessary. I’m not sure if there’s anything better.

JSON sure felt better than XML. But did we abstract too far, or did we just get so much better bandwidth and processors, that enriched data exploration is just insignificantly possible?

I read the use-cases for this, but I feel like: if you’ve gone so far to where this is helpful/useful, have you just gone too far?