Hacker News new | past | comments | ask | show | jobs | submit | igor47's comments login

Why are GCR and pgbouncer incompatible? Could you run a pgbouncer instance in GCR?

GCR assumes it's workload is HTTP, or a "job" (container that exits once it's task has completed). It scales on request volume and CPU, and the load balancer is integrated into the service. It's not obvious to me how you'd even run a "raw" TCP service like pgbouncer on it.

I’m not an expert, but from what I understand, the standard set up is like:

4x(Web processes) -> 1x(pgbouncer) -> database

This ensures that the pgbouncer instance is effectively multiplexing all the connections across your whole fleet.

In each individual web process, you can have another shared connection pool.

This is how we set it up


Cool! I actually did a similar project this year, though not expandable:

https://igor.moomers.org/posts/solar-ev-at-burning-man-2024


That's neat, though why do the two metal rails stick out over the windshield?

Also, dcsm is compelling.. are you located in The Bay by chance?

https://igor.moomers.org/posts/secrets-in-docker-compose


This is an over statement. I, too, have read Slouching Towards Utopia and Steven Pinker. But (a) technology is not a monolith, it's clearly nice to have enough food and modern medicine while it's also terrible that we're destroying our biome or facing increased illiteracy due to social media. Also (b) technology, like beer, is both the cause of AND the solution to all problems. Harari and James C. Scott make compelling arguments about how agriculture, among the first widely adopted technologies, deprived us of a utopia that we're scrambling to recover with more technology ever since. I call this the Operation Cat Drop problem. See also Under a White Sky.

[citation needed]


Curious how people interact with LLMs besides just going to chat.com/Claude directly. I've been trying aichat but not sure yet if it's worth it, especially given the token pricing vs the flat fee structure on the website.


I’ve found using AI tools strategically helps boost productivity. I use OpenAI Chat, Cursor, and Claude, each for specific purposes. Claude is great for coding without memory and serves as a nutrient-tracking diary, but I occasionally reset prompts to manage memory limitations. Cursor handles programming well but requires persistent, structured questioning. OpenAI Chat’s memory is a double-edged sword, useful yet needing occasional edits. Living in Poland, I rely on both Claude and OpenAI for translations, especially between English, Polish, and Ukrainian.

The humor comes when AI ‘misunderstands’—once, ChatGPT hilariously offered to vacuum a room instead of translating the words "Please vacuum the room”.



Why do you say that? Demographics is a fairly well understood field. Can you point to other moments when such forecasts have been wildly inaccurate?


So. 2100 (the end of year of the projection that peaks in 2080 and declines thereafter) is 77 years forward from last year, so we'd need to look at projections of current population from 1947 or so to gauge past similar projections.

The Census wasn't so bold then, but we can look at, say, the 1964 projections of population out to 2010 (only 46 years rather than 77 years out), which rather than a single projection had four “Series”, with the highest projection at 437,578,000 in 2010,and the lowest 321,916,000 and the midpoint between the two intermediate projections at about 375,000,000.

Actual US population in the 2010 census was 308,745,538. (And the projection missed significantly high despite all series using projections of increased life expectancy that were way below what was actually acheived and fixed immigration assumptions that were wat below the levels reached in even the 1970s, mich less the even higher levels reached later in the period; far more than 100% of the error was getting fertility estimates wrong.)

Now, a challenge for you: show a similar population projection to the 77 year Census estimate that proved reasonably accurate.


Black death? Post-WW2?


Rock Rabbit (https://rockrabbit.ai) | Full Stack Engineer | USA FULLY REMOTE | Full-time

Buildings account for ~40% of all emissions. There is a ton of federal, state, and local programs to help decarbonize buildings -- but this money is difficult to access, trapped behind complicated program requirements, uncertain eligibility, lack of clarity, and long payout cycles. We make access to these incentives easy by abstracting away the policy complexity and letting our users focus on their projects. Our platform serves contractors, homeowners, incentive providers, and equipment manufacturers.

We're hiring Full Stack Engineers for full time roles. We're well-funded, have a lot of customer traction, and are rapidly expanding the team. Our product vision is very clear -- the bottleneck is skillful, user-focused execution.

Tech stack: Python (FastAPI, pyright), Google cloud (GKE), React (Typescript, react-query, auto-gen API client)

We value curiosity, ownership and collaboration. Interest in energy, climate, and sustainability is a plus.

Apply at https://rock-rabbit.breezy.hr/p/6cbada45c1e5-full-stack-engi...


What does "USA FULLY REMOTE" mean? In the past, "fully remote" meant the job could be done from anywhere in the world, but adding the "USA" makes it sound like it's remote in the USA only, which is not "fully remote." It would be clearer to potential applicants to simply say "Remote in USA only".


Fully remote meaning there's no office and the company is 100% remote. The USA part means we only currently hire in the US.


no fully remote has always meant no office requirements


I wrote a little tool for this! https://github.com/igor47/dcsm


Fine nuance of exactly the type that's hard for folks to get behind today. If we could get people to consider nuanced positions, many of our problems would already be solved. Even in engineering orgs I've been a part of, navigating the challenges of seniority, hierarchy, and status has been really challenging.


I like using my blog to experiment with technology. I used to use GitHub pages and wrote some tips about using it with custom domains: https://igor.moomers.org/posts/github-pages-proxying-and-red...

But then I wanted to play with nextjs + typescript, have total control over how everything works, and host it myself in a container so I wrote a little static generator with next: https://github.com/igor47/blog

I think there are many ways to generate a site from images and markdown and the "best" depends on what you're trying to achieve


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: