Greg Kopyltsov

Father, founder, software engineer.

(VC Funded) - CTO & Co-founder
(Bootstrapped)
(Bootstrapped)

[Greg Kopyltsov @gregkop]

Seedream v4 just dropped on @fal, its a SoTA image generation model and looks like Bytedance's answer to Google's nano banana model

lets run through common prompts / use cases for these two models and see which one is better in which scenarios, left will be seedream and right will be nano banana

[Greg Kopyltsov @gregkop]

@shiri_shh Been playing with Muse 1.3 in OpenCode (it’s free currently) and it’s actually quite good, better than Gemini Flash 3.8

[Greg Kopyltsov @gregkop]

@BlendiByl H3 Max Turbo is fast but the infra around it is running kinda slow in my experiments

[Greg Kopyltsov @gregkop]

@dhruvbhatia0 these are the avg latencies I'm seeing playing with H3 Max Turbo - are ya'll having capacity struggles or is this a skill issue on my end?

[Greg Kopyltsov @gregkop]

@darrowoah What are average queue / latency times? I think FAL struggling to keep up with demand, startup times for H3 max there can spike to 50 seconds

[Greg Kopyltsov @gregkop]

@fal kind of defeats the point of H3 Max super fast generation in seconds if the startup / queue times can spike to 50 seconds - is this just a capacity thing from ya'll being slammed?

[Greg Kopyltsov @gregkop]

@matt_teeixeira then this morning must be painful for you as well https://t.co/hlPzouY4kq

[Greg Kopyltsov @gregkop]

@theaaron @grok @bot @NousResearch Why do you think Grok bot has had deeper market penetration in a shorter time than Hermes? Is it just the X marketing machine or that it’s more approachable?

[Greg Kopyltsov @gregkop]

@ryanvogel Now ship it as an Omarchy plugin!

[Greg Kopyltsov @gregkop]

@joenrv How’s the performance? Is it daily driver level or just a demo to get a feel for it?

[Greg Kopyltsov @gregkop]

@ericzakariasson @bot Better usage visibility, currently it’s easy to blow through the usage limits and hard to troubleshoot the problematic bot

[Greg Kopyltsov @gregkop]

I would recommend trying it, I imagine it’s going to be polarizing for users - I’ve been on Mac only for most of my career developing and Omarchy felt immediately snappier and more productive on a weaker / older machine

It’s also very satisfying to be able to change the OS on the fly by talking to an agent in my CLI

One example, I didn’t like the default screenshot image snapping so built my own plugin where it’s easy to share the image I snapped with a CLI agent after taking the screenshot

[Greg Kopyltsov @gregkop]

Found some time this weekend try @OmarchyLinux on an old desktop I had running Windows that was largely abandoned / just a plex server

First impression is that its...zen - that's the best way I can describe it, there is a satisfying power and simplicity

The OTHER cool thing is that its really brought this machine back to life, it actually feels great to develop on and is running a few coding agents and apps that my older 16GB Macbook would be choking on

[Greg Kopyltsov @gregkop]

@ky__zo Likely a BIN attack, bad actors find credit card forms to test large amounts of credit card numbers to see which are valid / invalid

[Greg Kopyltsov @gregkop]

@Daniel_Farinax I do wish there was better visibility from the Grok Bot team on at least which bots are burning through the most tokens - hard to optimize / manage the usage currently

[Greg Kopyltsov @gregkop]

have found a few useful patterns, one that I am liking today is connecting Granola and Linear and having it use granola customer feedback calls and pull out feedback around a particular feature or module and interview me about it for a Linear project - efficient way to turn client feedback into actionable issues ready to code & ship

[Greg Kopyltsov @gregkop]

@morganlinton I’ve been working with Grok bot to introspect and get it to give me some details around its internal working memory, so far seems like it’s markdown file based on its computer - curious how well this would work long term for you - have you considered external memory providers?

[Greg Kopyltsov @gregkop]

@tristanbob @bot I usually start with a “chief” or “boss” and then have a chat with it about what we’ll be working on and which sub agents it thinks we should “hire”

Then the boss can make those grok bots for you and even group them appropriately so that they collaborate

[Greg Kopyltsov @gregkop]

Has anyone had any success adding custom MCP or plugins to Grok Bot? want to add one to help find inefficiencies in context sharing between bot teams that burn too many tokens but seems impossible currently

[Greg Kopyltsov @gregkop]

@notrab Came to see if it’s better than orbstack

[Greg Kopyltsov @gregkop]

Was looking at the @fal example for mirage-api/avatar-x/reference-to-video - is it just me or does the neck movement look really really weird?