nijave

11 ms·

241.
by nijave 4mo ago
Yeah slowly building up to that, we noped out of DD APM because leadership does not like seeing the $ amount of the bill (despite it being a small fraction of AWS and consisting of 10s of services in 1 platform)
242.
by nijave 4mo ago
LLMs are an interesting call out. I've been trying with the idea of a query optimizer agent that analyzes PG slow query/auto explain logs. I've only tried Claude Opus but it does a pretty good job interpreting the plan. It&#x
243.
by nijave 4mo ago
Semi related, friend works here https://www.nextdata.com/product (looks like they rebranded for AI heavily...) AI buzz aside, unified multimodal data platform is an interesting idea.
244.
by nijave 4mo ago
What are the problems you have with tooling? Imo it's no worse than most other languages besides a very small handful of recent ones (rust, go) where everything is included The easy approach is usually just throw it in an OCI container
245.
by nijave 4mo ago
Because the English to code translation step is fallible
246.
by nijave 4mo ago
Anything performance sensitive ends up being an extension in compiled code anyway Python is mostly just glue
247.
by nijave 4mo ago
>In about 22 years of writing ruby code, I have never ran into a situation once where I would have caught a bug through types. I've definitely ran into that although much less common at places with good test discipline. I think the
248.
by nijave 4mo ago
Not necessarily. You can do async replication either at the app level or DB level to other regions Each individual user is fast due to close geo and everyone else has a small (potentially trivial) lag to see writes. Not sure if such an arch
249.
by nijave 5mo ago
The parent is about vendoring dependencies which Nix does not do by default so that is not a solution here.
250.
by nijave 5mo ago
Yeah imo that is bad advice. In my experience, lockfiles do as designed and exact pinning in the high-level manifest makes it extremely hard to do periodic updates because you end up spending hours tinkering with pins to try to find the rig
251.
by nijave 5mo ago
Caching proxies are a decent middle ground like Artifactory. AWS might support that (maybe only on certain repo types?) Generally you can also configure rules in your internal package cache about what to do if a package is missing from the
252.
by nijave 5mo ago
> :black_circle_for_record: Smoking gun. > "belt and suspenders"
253.
by nijave 5mo ago
I imagine some scenarios where you're throwing a big data set at an LLM could add up quickly. "Check all of Confluence for outdated and conflicting info" "Review all the legal contracts over the past 10 years" &quot
254.
by nijave 5mo ago
In the corporate world, writing code was always a small chunk anyway. Iirc something like 30% of employee time. Getting 50% faster there still only gains back 10% of your time. To further complicate matters, you had to spend on AI software
255.
by nijave 5mo ago
Someone at work made a Claude skill that seems to work, surprisingly
256.
by nijave 5mo ago
Unless you're on sqlite, the database is still an external dependency with network requests
257.
by nijave 5mo ago
Really any "constant". Failure thresholds, timeouts, API versions or endpoints, LLM model id
258.
by nijave 5mo ago
Fwiw I didn't think Cloudflare sales was too bad when we dealt with them a year ago--at least compared to some companies.
259.
by nijave 5mo ago
Haven't used Vercel but back in the Heroku/CloudFoundry days it was pretty easy to jam arbitrary binaries into the runtime containers and some of the build packs were flexible enough you could override most of the functionality. N
260.
by nijave 5mo ago
I think they were doing that before AI got big the last couple years. Their core network stuff always seemed pretty robust but all the newer stuff was much more thrown together. Thinking specifically of Zero Trust/Argo Tunnels which ha
261.
by nijave 5mo ago
Git is typically fairly slow if you have to wait on a test suite and deployment pipeline. Usually at least 10 minutes but sometimes 30, 60, 90+ minutes. A lot of purprose-built feature flag platforms hot reload the config in seconds. JSON i
262.
by nijave 5mo ago
Assuming you want a random distribution and don't want to take any other attributes into account. We're a small company but new feature release for big features is typically targeted at low risk users/customers first. That us
263.
by nijave 5mo ago
Hmm, I have some anecdotal evidence this is true. Interactively working out a plan with Opus on multiple occasions it'd come up with an incompatible solution, I'll add additional context/requirements, and it has a tendency to
264.
by nijave 5mo ago
Switched back when 4.7 had an issue last week and it was wayyy faster. I assume mostly because a lot of people have moved off but might consider using it more just for the speed boost.
265.
by nijave 5mo ago
https://www.protondb.com/
266.
by nijave 5mo ago
dd filesystem to another machine then boot it up with an emulator like qemu and do a trial run Be careful if you have anything that autostarts that reaches out
267.
by nijave 5mo ago
Curious what cases you'd want this that IMAP+SMTP or email MCP don't already solve
268.
by nijave 5mo ago
IMAP?
269.
by nijave 5mo ago
But that doesn't help for the agent receiving mail from arbitrary 3rd parties
270.
by nijave 5mo ago
Ironically they could have used those AWS keys to use one of the many AWS services that's more secure. For example S3 (ideally with KMS), Parameter Store (ideally with KMS), EBS, EFS, AWS Secrets Manager, even just KMS to directly encr

More ›