kndwin

@MichaelArnaldi Keen, been building with effect full stack with effect-atom and it's the most asynchronousable stack

Effect will become a full stack framework including functionality of next/tanstack/etc. Other frameworks are nice but they fail the test of AI and integration is poor. End to end type safety requires design from the ground up, and no end to end inference isn’t type safety.

@BatsouElef It works as of now, polishing the edges and adding stuff as I see fit. Focusing on polishing and documenting capabilities (chat is the main one but there’s more like task and repo management)

@kndwindev Is this done, or are you planning to add more things?

Share your website(s). 🔗 Last time it reached 32K views! Describe in a few words what your product is solving.

@devagrawal09 If your data can be scoped to an org or a workspace then it’s a good candidate for sync This is because syncing is harder when you have to worry about auth (what do you sync) If all the data is authenticated by nature then you can just sync.

@kndwindev what do you mean "workspace your data"

What is your mental model of a "sync engine" vs "data fetching and mutation in a web app" Who should be using a sync engine and why? Who should not be using a sync engine?

@jacobmparis I've also been trying to build a github clone. Curious if you guys also built the ssh server to push as well or building the skin with github servers. Either way, finding a stable caching strategy has been painful, I have something "working" but it's mad slow on the reads

GitHub is actually a very hard web app - heavy rate limits - big data fetches - either get slow nav or slow initial load until cached - but you can’t share cache for private repos - must purge cache when removed from repo I have a POC in Next but missing one feature

@ohmypy I think they keep trading off "get it working" and don't refactor to "make it simple". Even with AI, the dopamine (and reward) of getting it working good enough is easier than taking the pain to re-write it again but simpler. Sucks though cause simple is mastery.

Here's what I don't get. Why do codebases degrade so quickly, even on brand new projects, and turn into an unmanageable mess in just a few months? Why don't developers care more about maintainability and design at the package level?

@ykgup @rauchg It’s not trivial though, you have to install bolt server and use a JSON based UI. Doable for a bigger org sure, but annoying for something that should be much easier. Building qave.chat to combat that, building this with precise intent using AI

@rauchg the Slack as platform angle is underrated. most companies are already running half their ops through Slack bots. making that layer programmable with real infrastructure behind it is a bigger deal than it looks

"Creation as a Service" will always beat "Software as a Service". I don't want your lowest-common-denominator software. I want software to work for me and meet my needs. I want platforms and infrastructure. To me, Slack is a good example of an ideal in-between of SaaS and Platform. I have zero interest in vibe coding my own Slack. (It's also particularly difficult software to perfect and scale, and if you disagree you have no idea how to build software 🤡). I do have a strong interest in extending Slack with our own agents. As I shared yesterday, more of our company runs on Slack-based "@" agents. We're making it easier than ever to deploy Slack agents on Vercel, with our AI infrastructure (gateway, workflows, sandbox, etc), starting with @v0. A lot of SaaS might be cooked, but Salesforce has a generational re-birth opportunity with Slack.

@awsdevelopers AWS

Reply to this tweet with "AWS" and we’ll tell you which AWS Service you are

Fair enough, I do think Effect maintains organic complexity really well. I have introduced Errors as value into prod before and have felt the pain of it. The precision was too verbose. Effect gave me flexibility at handling errors at any layer but it was hard to get adoption due to fundamental shift in primitive. My 2 cents is most software I've encountered has handled their organic complexity in a need by need basis and Effect makes it hard to incrementally adopt (I know you can, but have 2 distinct patterns of doing thing is hard so it's usually a hard swap or you blackbox it). The patterns of Effect are powerful, I wish I could steal some of it and use it at work where the decision of how things are done were made prior to me being there. Which is why I think this is a good step towards it.

@kndwindev that's simply because you don't realise the investment it takes to adopt the other, you don't need to make a counter argument, use it at scale and you'll realise the difference. Error as values is nice for example code, it breaks hard when you use it

You’ve done the 2% now do the 98% and let’s see

@__morse I like this, Effect was too hard to sell to my team. This feels more incremental and easier to adtop

here is the same Effect code but using errore.org philosophy instead. much easier to understand. zero dependencies.

@harjotsgill Context plumbing has high friction with authentication. You can't access the tools you need unless you authenticate. But instead of resigning to only local, why not just auth once and have all the tools you'll ever need? that's the end game of qave.chat

OpenClaw worked because it had access to all the integrations locally, which is why Claude Code CLI also succeeded - the code is available locally. In contrast, background agents like Codex/Cursor cloud (including Devin), ChatGPT apps, and others haven't been as effective, mainly due to the friction involved in connecting GitHub and other services. Most well funded players focused a lot on cloud agents vs release scrappy locally run agents (which are harder to monetize). This is also why new products are now focusing on a local-centric approach, with agents such as the Codex desktop app running local automations using cron jobs, the Claude desktop app with SSH access, and etc.

If you’re an engineer building product. You’re in a unique spot where you can test UX intuition without being blind to the technical debt and tradeoff. Be very ambitious on building the best UX. It’s much more fun

@neetcode1 @thdxr @opencode Would love on how Dax builds software primitives. The software can that he’s produced are high quality in surface area and wide in scope

I'll be having a conversation with @thdxr from @opencode - what should we talk about? Gimme your questions.

@thdxr OpenCode and STS are the golden standard of how I hope to write software. My gut is even with the forks, the intuition of how you guys build will seem to out perform these competitors.

lot of commercial OpenCode forks floating around that we get tagged on so gonna write up our positioning OpenCode is very intentionally MIT licensed - we hope to do such a good job that companies building coding agents opt to build on top instead of reinventing the undifferentiated parts it's tedious - getting things working across hundreds of LLM providers, dealing with quirks, collaborating with all the companies involved to improve things and then the platform work to get things working well on all environments thankfully we have a solid business model and can devote significant resources to doing all of this - including funding things like OpenTUI the forks that have popped up are confirmation we're starting to hit these goals so that's great - even when they're from competitors what's less great is they seem bent on passing off all this work as their own, even claiming they "built it from scratch" they've never communicated with us, never privately said "hey thanks for the work" every company can choose how they want to operate but this isn't very aligned with the spirit of open source and these small gestures of good will go a long way

Token output being slow has forced me to parallelise my workflow. Before I would go deep on one task, wrap it up and do the next. Now I'm checkpointing hard tasks more often for the sake of parallelisation. Definitely feeling the context switch tax, but it's an interesting skill to practice. I suspect as models get released I'll switch from one to the other. Codex Spark will switch me back to deep flows quickly, and the next frontier model will switch be back to parallelise.

@thdxr I usually do, especially if the tweet has variance of interpretation that I would like to see.

does anyone read replies anymore? i'm assuming the bots killed this

@suni_code qave.chat

Drop your SaaS, I will use it and give my feedback 👇

@1Umairshaikh Been building qave.chat on the side, having a lot of fun! Recently added a CLI to support getting stuff `bunx @qave/cli@latest` Try it out!

What are you building on weekends? Share links 🚀