doh

I built a very similar server myself [0] with a similar setup. I run different models for different purposes, but the primary one currently is kimi 2.6. I run kimi as the orchestrator model and then qwen, Gemma and others for specific tasks (sometimes loaded dynamically based on the task at hand), all exposed through the pi harness. I also use Hermes for some personal repeated tasks which connects to the same models, hosted on my local Mac Studio.

I am not even going to pretend that this is financially reasonable option. I simply wanted to have a local models. Maybe down the line, as cloud models become less subsidized, I might benefit from having a local setup, but for now, it wasn't the most prudent financial decision.

But one big benefit is that I never have worry about my account being randomly banned nor I have to worry about running out of quota. I still use codex and opus for some specific tasks, but as tools are improving, I need them less and less.

[0] https://x.com/synopsi/status/2024235558193811778?s=20

I have built something like this and in process of collecting the data.

Frontier users: 527,865 Light indexed: 527,865 Ready to queue: 9,083 Fast scores ready: 0 Activity events 24h: 30,266 Fast scores completed 24h: 19,123 Deep jobs completed 24h: 3,043 Fast-score ETA: n/a Deep-hydrate ETA: 69h Stale running jobs: 0 GitHub backpressure jobs: 19,113 High automation signals: 4,608 Medium automation signals: 1,327 Completed jobs: 74,714

Biggest challenge is Github's rate limits. At this pace it will take two more months to have 98% coverage. But after that the maintenance should be quite straight forward.

I don't want to defend LLM written code, but this is true regardless if code is written by a person or a machine. There are engineers that will put the time to learn and optimize their code for performance and focus on security and there are others that won't. That has nothing to do with AI writing code. There is a reason why most software is so buggy and all software has identified security vulnerabilities, regardless of who wrote it.

I remember how website security was before frameworks like Django and ROR added default security features. I think we will see something similar with coding agents, that just will run skills/checks/mcps/... that focus have performance, security, resource management, ... built in.

I have done this myself. For all apps I build I have linters, static code analyzers, etc running at the end of each session. It's cheapest default in a very strict mode. Cleans up most of the obvious stuff almost for free.

I am in SD and would love an invite. I am keep thinking about uniting more like minded people for a while. My email is r@seslu.com

Hear hear

Pex | REMOTE (Europe), FULL-TIME | https://pex.com/careers/

Pex is a digital rights technology company, enabling the fair and transparent use of copyrighted content at the speed and scale of the Internet. We serve everyone who uses the Internet to view, share or create content – from the largest platforms and rightsholders, to independent creators. Our advanced licensing infrastructure allows platforms to manage and license content before it’s published, empowering creators to upload freely while respecting copyright. In return, rightsholders are able to monitor and capitalize on the content they own.

We are hiring for Senior Reverse Engineer/Senior Data Harvesting Engineer (Europe) to help us bring our services to the masses! Reach out to jobs@pex.com.

We offer:

  - salary: $103,000 – $110,000 USD per annum
  - equity, with a 10-year exercise window
  - 30 days of paid time off + 9 local holidays + the day off on your birthday
  - generous paid parental leave
  - a fully remote work environment, supportive culture, and excellent work-life balance

  - Interviewing at Pex? Here's what you can expect during our hiring process [1]
  - Pex culture: Focusing on what really matters [2]

  [1] https://pex.com/blog/interviewing-at-pex-heres-what-you-can-expect-hiring-process/
  [2] https://pex.com/blog/pex-culture-focusing-on-what-really-matters/

There are 27 countries in EU with different motives, morals, interests, etc. Just because you agree with the decision of one country in one instance it doesn't mean you would agree with them all. But once you give them the powers it's impossible to take them back. It's a bad slippery slope.

The courts are altering their views on scrapping. This [0] is a good paper that explored the last 20 years of rulings (although it hasn't been updated with the most recent cases).

[0] https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3221625

I am curious if you would be interested to retry the idea? I might have an in with the YouTube team. I feel like it's a shame to let this go. Would you be open to chat? Please reach out r@pehul.com

Quick note for the osohq team: The "Read the docs" button leads to 404

Modus [himodus.com] | REMOTE (US Only) | Full-time | Staff/principal positions only | $300k + 5% stock

Modus is a continuous workforce management platform that makes headcount planning and reconciliation easy. The platform analyzes thousands of bits of data across the organization to help leaders optimize their workforce and increase forecast accuracy. With the help of AI, Modus synthesizes insights from HRIS, ATS, expense and finance systems so leaders can visualize and execute workforce plans, grow efficiently and compliantly, and unlock high performing teams in one unified workflow.

While we are just at the beginning, both founders have extensive experience in the industry. One of the co-founders was early Google engineer and built multiple companies in the past, writing significant amount of code at each company. The other cofounder has scaled from 100 to 2500+, through to IPO and beyond. Joining as our first engineer and designer means you have will have an outsized influence on the tech and design choices.

We are currently only looking for very senior colleagues to help us move much faster:

   - Staff level frontend engineer - At least 5 years of staff level frontend development experience with Svelte, React, Next.js, Vue (strong preference for Svelte)
   - Principal Product Designer - Preferably designer with strong rapid prototyping skills, 10+ years of experience with designing B2B software, especially complex dashboards with a lots of data (strong preference for either Fintech or HR tech experience)

Interview process: phone call, interview with each founder separately, reference checks, and offer (can be done in less than a week).

If interested, please reach out to hire@himodus.com.

I use similar pattern myself. Was curious if the OP is using some other, like for instance splitting the struct into two (im/mutable) and then passing them around, or what.

BTW kudos on zanzibar. Love the tech and the code).

I think that's a valid criticism. What do you think would be a more ergonomic pattern?

Honestly someone who understands the environment they are developing in. I don't care about the language as much as I care about understanding the choice they make.

But ultimately we are using SQL, queue, concurrency, single tenancy and monolith (few to none micro services).

I understand your sentiment and makes sense.

Unfortunately, majority of candidates are interested in the take home challenge over doing live coding session. And we optimize for the widest reach.

In regards of the results, I can't speak for others, but I write code for 26 years. What I look for is how the candidate thinks of the problem. Here is one of my favorite exercises [0]. I would obviously love to do this live but people get super nervous and stop thinking. I then have tendency to help them too much and learn much about them.

[0] https://news.ycombinator.com/item?id=39263923

Let me give you an exercise I really like for this.

Imagine a file with endless list of URLs. For each URL check if it contains an image. If it does, identify the top 6 colors and store it memory so it can be printed/stored later.

In something like python this can be as short as 20 lines of code: - open the file - read line by line - validate if the URL is valid - call the URL - check if it's image - if image, read pixel by pixel and check the color - store it in dictionary/hashmap

Here is what I look for: - does it compile/work? - what language did you choose? (like python is fast to write this in few lines but hard to introduce concurrency) - have you introduced some kind of caching for the URLs? (so you don't waste resources) - are you checking if the URL is valid or you just run them all and wait for errors/timeouts? - are you checking the file type from HTTP header or body of the file? - how do you handle errors? - are you attempting concurrency? If yes, how?

I actually done this exercise myself in dozen or so languages to see what choices I make based on the language. Obviously one can spend significantly more time on it if they really desire, but the design choices are visible from the beginning.

I very much understand your sentiment. Unfortunately we haven't found any other functional way to achieve the same goal.

I am not sure about other firms, but we personally look for how the candidate thinks about the problem more than anything. Did they handle exceptions well. Are they focusing on efficiency, readability, speed, or?

We found out that from a simple assignment that shouldn't take more than 30 min, we learn more than sitting with the person for hours. Plus they can do at their own pace without someone breathing down on their neck.

Modus [himodus.com] | United States | REMOTE | Full-time | $150-220k + very early shares, employee #1-5

Modus is a continuous workforce compliance and planning platform that enables various employees throughout the organization to stay compliant with labor laws, visualize and execute workforce plans, and make data-driven decisions to ensure fairness and equity. You can think of us as Vanta for HR. Our mission is integrity. We've all been through the pains of levels, benchmarks, compensation inequities. Why can't HR tech be intuitive, easy to use, and encompass workflows for everyone in the org?

While we are just at the beginning, both founders have an extensive experience in the industry. One of the co-founders was early Google engineer and built multiple companies in the past, writing significant amount of code at each company. Joining early means you have a significant say into tech and design choices.

We are looking for curious and motivated colleagues to turn our vision into a reality. All positions are 100% remote with no on-site requirements of any kind. We are currently looking for:

  - Javascript/frontend developers
  - Designers
  - Backend developers

If interested, please reach out to hire@himodus.com

I heard this story first as a kid, but with a wheelbarrow.

I do agree with this with the caveat that EVs are heavier than ICE. So it could be weight + emission = tax. That way while EVs are heavier they have no emissions and thus the price equalizes at some point.

This is why I pay almost exclusively with Amex. They do contact a vendor and sometimes the vendor tries to fight it, but if I reopen the dispute, I always get my money back (and I dispute a lot of transactions).

I have the opposite experience with Mastercard. Of the dozen or so disputes across 5 cards I had over the years I essentially lost them all. Visa is somewhere in the middle. They tend to side with customer but also will investigate thoroughly.

Well, fuck.ing costs only $12,999.99/yr.

Can you though? NordVPN for instance comes to $67.5 for the first year and then $120 every year. I know there are cheaper options too, but not that much. Plus they still have to buy the subscription. And then dealing with the hassle of the VPN.

Super cool. Thanks for the suggestion found this [0] but there appears to be more out there.

[0] https://www.hcii.cmu.edu/news/software-defined-cooking-using...

Thanks for putting me on this topic. Will do much more research.

I keep thinking about this topic for a couple of years now (hence my question). I wonder if combination of new technology could be made into much more useful microwave, even if the cost was not suitable for personal use (at least initially).

The dream is a microwave with "single button" (start/stop) and a temperature scale (lukewarm, warm, hot, boiling) that consistently heats up food correctly and doing so within very short period of time (below 3 min).

The issue is time. They absolutely work better than microwaves, but microwaves are popular precisely because they are just much faster.

I am not expert, but my understanding of cyclonic inverters is that they are able to lower power output and thus achieving more even cooking. That also implies slower time. I am more interested in precision while retaining the speed. But it's something I looked into myself too.

Thanks for the context. That seems to invalidate the idea.

Rather than heating food more or less, I care about equally. Couldn’t combination of thermal camera and ultrasonic transducer achieve the goal?

If they indeed need a contact with the object, then obviously that wouldn’t work. But I am not sure why the air molecules couldn’t carry it.