Nicolas Alpi 🍪

Do you use Twitter likes / Medium recommend / Github favs as a form of bookmarking tool?

For whatever reason, it seems that the X polls won't work :/ But would love to hear people's thoughts!

I have a serious question for software developers. Help me reach as many people as possible 🙏 If you're using any AI coding tools regularly/daily right now (Claude Code, Codex, Cursor, Gemini CLI whatever). How do they make you feel as a developer? * Super Human*: I can build as many things as possible so fast 💨 *Focussed*: Speed it not my priority, but I get to achieve more focus by having an AI thing working alongside me all day 🧘‍♀️ *Lazy*: Why do I even need to work, when a black box can write my code while I warm up another coffee 🥱 *Frustrated*: Yea, I guess it's cool, but doesn't this thing can fix that very simple bug!!! 🤬

🚨 Major active supply chain attack just hit npm. Popular package @​ctrl/tinycolor was trojanized — and it didn’t stop there. Over 40 packages were silently modified to steal secrets from dev machines & CI pipelines. Our team at Socket caught it. Full report coming soon. Stay safe out there.

Now the next step will be how do we as CTO/Leaders share these with our teams so we can have a unified AI augmented experience. Long story short, this space is actually quite exciting. Turns out, it's not a fad 🤣

So yea, our job as CTOs / Senior developers are rapidly changing. Does it mean it will be less enjoyable coding tasks, I don't know. I seem to still get my coding fix when the tools can't achieve something or when I simply don't want to use them. But, there's certainly a new area to learn. I doesn't matter the tool you use, or even if you use those tool, learning how to specify a problem, and breakdown a solution is at the root at what we do as developers. Writing the code should only be the last part of the job.

**Senior software developer 👴 VS AI 🤖 (a 🧵)** I've been developing software professionally for 25+ years. I'm not a CTO but still very much enjoy coding and problem solving. I've worked with, mentored, and up-skilled a ton of more junior developers over my time in the industry.

I love the apple keyboard weirdly, and the only keyboard that replaces it for me is the Typematrix because of the vertical layout and the centered return and back key. Super confortable for development sessions. Unfortunately I think they stopped production :/ But I have at 3 or 4 here so 🤣

You switched from the Apple Keyboard to something better. Which one and why do you like it better?

Alright, I officially have a #RailsWorld FOMO now :/

Our biggest reservation in using Kamal is "what to do when things go wrong". Non of us has enough knowledge in Docker to really feel confident we can debug in a decent amount of time for a production app. Wonder if there's something here to add to the course? (what's likely to go wrong, what to do, where to search for answers etc)

Kamal DevOps course in a tweet Terraform - optional Kamal DockerHub Cloudflare - optional Tailscale - optional 1Password - optional Hetzner SQLite (with Litestream) PostgreSQL Prometheus - optional Loki - optional Grafana - optional

@elliotkoss @greg_blass @jonathandavies @heroku @render You would setup replication first between the Heroku DB and whatever external DB you want. And once the replication has caught up, can handle the switch. Ours is a bit more complex because we're behind Shield.

@nicalpi @greg_blass @jonathandavies @heroku @render I'm aligned with you. Where I struggle is how to ensure data redundancy prior to a switch over (or how to handle a few hours delay for active end users), the escalation path to knowing when to engage the switch over, and the resolution path. And is this effort worth the cost.

Do you know what, outages happen. And I feel for the @heroku devops right now. But, the lack of communication is a real killer here.

I guess, rn my thinking is that it should be ok to accept some level of single of failure really at our scale (after all, our DNSs can also fail). Maybe a multi region replicated RDS with two different PAAS providers that connect to it, where one is more on a standy / costing almost nothing but still being deployed on every CI and tested time to time on stagings. If something happens on one, we can switch at DNS level to the second one kind of thing. I haven't though that fully yet, but I guess it's a path that should be investigated if we want to continue to use PAASs.

@nicalpi @greg_blass @jonathandavies @heroku @render The data back-up is something that I'm wondering as well. I'd love to hear the path you wind up going down.

@jonathandenney @zenoix9 @bkessler @raoulsnotes @jessethanley @RylanBlowers @thedomcs @fredrivett @heroku Did that on the web ones yes, and they reprovisioned, and so far the web is holding up quite nicely.

@nicalpi @zenoix9 @bkessler @raoulsnotes @jessethanley @RylanBlowers @thedomcs @fredrivett @heroku So you're running heroku ps:stop worker.1 once they stop?

Is @herokustatus @heroku having a major outage right now?

@jonathandenney @zenoix9 @bkessler @raoulsnotes @jessethanley @RylanBlowers @thedomcs @fredrivett @heroku What *seems* to have worked for us was to scale up (and eventually get it to work) and stop any previous Dynos (which led to reprovisioning). Now the web part seems to at least be back up.

@zenoix9 @bkessler @raoulsnotes @jessethanley @RylanBlowers @thedomcs @fredrivett @heroku It appears there's some delay in the dyno lifecycle right now, even when getting successfully scaled messages. When the running dynos eventually shut down, I'm scaling down and then scaling up to cycle to new dynos