beoberha

Hope you never want to take diff backups

I believe OP is talking about new sessions or after compaction. He’s getting at the fact that LLMs are stateless and have to rediscover your codebase on every new session.

For me, the concern about SQLite has never been if the database engine itself is “reliable for real data”, but that storing data on a single node is not “reliable for real data”. Performance aside, what you are positing is no different than dumping everything to a text file on disk. What happens if that VM dies?

There’s no better option today because it’s impossible to make it a better experience. That machine at home will need upgrades, it could fail, it costs thousands, it sucks lots of power. There is no mass market appeal.

What you’re describing is more likely to manifest as a proprietary product from someone like Samsung or Ring (likely both!) than an open standard AI server that integrates with everything in your home automatically. This is exactly like what we have today with security systems and smart appliances. You have managed services and you have Home Assistant in your homelab.

I don’t think there’s anything different between what you’re suggesting and a homelab. Most people do not have a homelab and are happy to offload services like photo storage or security to remote providers.

Spend 5 minutes looking up how to make a chat bot and be amazed how Telegram is really the only option. I was dumbfounded when rolling my own agent.

iMessage is proprietary. WhatsApp charges you. Unofficial APIs exist, sure, but not my cup of tea.

Then you have Discord or Slack, which are pretty heavyweight when all you want is a simple chat interface.

Telegram makes it SO easy. Bots are first class resources on Telegram and they make them so easy to use.

Bun is a full fledged JavaScript runtime! Think node.js but fast

Sounds like OP did some pretty cool engineering to make this run performantly. Definitely not your run of the mill AI slop.

Seems like a great fit - kinda surprised it didn’t happen sooner. I think we are deep in the valley of local AI, but I’d be willing to bet it breaks out in the next 2-3 years. Here’s hoping!

who knows when Apple decides to enter the game, but they will absolutely crush the personal agent market when they do.

My pessimism is mostly rooted in the VC economics of it all. The vision is great, but its a busy space and there's no actual product or business. They basically wrote the guy a check to build the spaceship in space.

Ignoring the VC economics and awful name, I won’t be as pessimistic as everyone. I see the vision.

That said, nobody knows what the AI future looks like. Entire’s entire thesis is a solution for something we don’t even know we need. It’s a massive bet and uphill battle. Traditionally, dev tool success stories come from grassroots projects of developers solving their own problems and not massive VC funded efforts that tell you what you need to do.

The initial excitement of LLMs has significantly cooled off, the model releases show rapidly diminishing returns if not outright equilibrium and the only vibe-coded software project I've seen get any actual public use is Claude Code, which is riddled with embarrassing bugs its own developers have publicly given up on fixing. The only thing I see approaching any kind of singularity is the hype.

I am absolutely baffled by this take. I work in an objectively high stakes environment (Big 3 cloud database provider) and we are finally (post Opus 4.5) seeing the models and tools become good enough to drive the vast majority of our coding work. Devops and livesite is a harder problem, but even there we see very promising results.

I was a skeptic too. I was decently vocal about AI working for single devs but could never scale to large, critical enterprise codebases and systems. I was very wrong.

Your sentiment resonates with me a lot. I wonder what we’ll consider the inflection point 10 years from now. It seemed like the zeitgeist was screaming about scaling limits and running out of training data, then we got Claude code, sonnet 4.5, then Opus 4.5 and no ones looked back since.

That’s kind of my point. They’re not really in competition. I bet they’d have an easier time with this scale if they were on SQL Server, but obviously that migration isn’t happening and startups don’t reach for it for many reasons.

Are you saying this because OpenAI didnt choose SQL Server?

I had never written an iOS app until a couple months ago and was initially very put off when I hit the same wall. The alternative is to host on a cheap VPS and find some way to prevent other people from using your app. When you cost it out, it's close enough to the 100 bucks a year for the Apple account. However, the kicker for me is the side loading process. Way too much headache compared to a deploy script that has my changes running nearly instantly.

Beads is amazing. It’s such a simple concept but elevates agentic coding to another levels

Yep, as a manager, I am explaining this conundrum often. You can be a rockstar SDE 2 or senior, but not be ready for a promotion because you aren’t leading enough.

I have a feeling most of these folks are talking about personal projects or work on relatively small products. I have a good amount of personal projects that I haven’t written a line of code for. After bootstrapping an MVP, I can almost entirely drive by having Claude pick up GitHub issues. They’re small codebases though.

My day job is mostly a gigantic codebases that seem to still choke the best models. Also there’s zero way I’d be allowed to tailscale to my work computer from my phone.

It’s a fast moving field. People aren’t coming up with new ideas to be performative. They see issues with the state of the art and make something that may or may not advance things forward. MCP is huge for getting agents to do things in the “real world”. However, it’s costly! Skills is a cheap way to fill that gap for many cases. People are finding immediate value in both of these. Try not to be so pessimistic.

I’m incredibly biased (I work at Microsoft) but I love Teams. It’s a great meeting app and a great chat app. It blows my mind that there are companies that have totally separate apps for each (Zoom/Slack).

I definitely agree with you, but it’s probably a little apples and oranges. MCP server is a one stop shop for discovering “tools”. To leverage a CLI tool “from scratch”, your agent has to do a web search to find if a CLI tool even exists, figure out how to install it, install it. Not saying those are impossible, but it’s way less automated and “deterministic” than what MCP provides.

Thanks! Didn’t scroll down far enough

Curious why pgduck_server is a totally separate process?

Sam is a great twitter follow

A lot of reductive anti-cloud stuff gets posted here, but this might be the granddaddy of them all.

Former players being hired as head coaches without much other coaching experience is fairly common in the NBA.

I’m sure the sensor signals look different when you press a button vs when you hit a wall. Enough that it could certainly be learned by a model