Ryan Dunn

I still build cloud stuff and brew beer.

[Ryan Dunn @dunnry]

I'm at Bluesky. Fuck this place.

[Ryan Dunn @dunnry]

It's a Hermes cron job that worked for months. Only deepseek flash fails. Other models work fine.

[Ryan Dunn @dunnry]

@opencode deepseek has issues: RuntimeError: Error code: 400 - {'error': {'message': 'Error from provider (DeepSeek): Failed to deserialize the JSON body into the target type: tools[0].function: missing field name at line 1 column 132', 'type': 'invalid_request_error', 'param':

[Ryan Dunn @dunnry]

Having lots of issues with OpenCode Go plan recently. It thinks deepseek-v4-pro is not there. cc @thdxr

[Ryan Dunn @dunnry]

@Aaronontheweb Ugh. I hate FRA - the absolute worst for connections. The amount of times I have been literally running to the furthest Z gate after it took 1 hr to get off plane, take bus, go through customs, security, then haul ass 30 mins away to furthest terminal ever. Hate is mild.

[Ryan Dunn @dunnry]

@Aaronontheweb Yeah, that's a good one. It's fast. I really like Qwen 3.6 27B as well - it's not nearly as fast, but it's smarter than 35b.

[Ryan Dunn @dunnry]

@cjzafir I have all these things. It would be interesting if you published a step by step with example prompts. Would love to try it myself.

[Ryan Dunn @dunnry]

@angel_d_munoz Look at OpenCode Go plan for $10/mo. Has all those models plus DeepSeek Pro and Flash.

[Ryan Dunn @dunnry]

I worked with him back in the day. He was a really smart and compassionate guy. Sad to hear this.

[Mat Velloso @matvelloso]

Shocking, and heartbreaking news

https://t.co/53hsfcg0Cx

[Ryan Dunn @dunnry]

Running into a lot of 429 errors using Hermes on @opencode Go plan with DeepSeek Pro. Seems fine on other models like DS Flash. @NousResearch anyone else reporting this?

[Ryan Dunn @dunnry]

@angel_d_munoz Deepseek 4 Pro and Flash have been pretty good. Not quite as good as GPT 5.4 or 5.5, but quite capable. Flash does a good job if it is guided by a better model.

[Ryan Dunn @dunnry]

@ivanfioravanti I have a 128GB M5 Mac and I wasn't able to get this to load. Do you have a pointer to which model you used on HF and which server? I tried LM Studio with https://t.co/W9H9pMigaV and it wouldn't load in LM Studio.

[Ryan Dunn @dunnry]

I remember when SkyDrive used to have a nice remote access client so you could connect to any of your machines remotely. I miss that still.

[Omar Shahine @OmarShahine]

April 2012: SkyDrive (now OneDrive) made the leap from cloud storage to true cloud sync, with desktop apps for Windows and Mac, mobile updates, 7 GB free storage, and a 25 GB upgrade for existing users.

A milestone worth putting on a cake. 🎂☁️

[Ryan Dunn @dunnry]

@peduarte @raycast Something's off

[Ryan Dunn @dunnry]

@LyalinDotCom I have the new M5 Max 128 now. What's your research and recommendations for local models and harnesses now?

[Ryan Dunn @dunnry]

Yep, that tracks. My teachers brought me and classmates to the library to traumatize us by watching it on replay. They did a great job of pretending to not know what happened.

[Dave W Plummer @davepl1968]

99.9% of people who "experienced" the Challenger disaster saw it on replay and now remember it as live.

Almost NO ONE was watching. Everyone thinks they were. It's a fascinating collective false memory.

[Kari Michael Hagemeier @kariha]

Just open-sourced my dotfiles — 100% built by AI agents 🤖

One command bootstraps a complete dev environment on Linux, macOS, Windows, or WSL.

📖 https://t.co/hHatMOGkeX 💻 https://t.co/iloFYrjyim

[Ryan Dunn @dunnry]

@rivet_dev looks really interesting. But, I don't get how the SQLite instances are stored/backed. To self-host this seems to require just compute and PostgresQL.

Where are the SQLite instances? Serialized and stored in PostgresQL when your actor is scaled down? @NathanFlurry?

[Ryan Dunn @dunnry]

@Aaronontheweb @dariogriffo I have 12 GB 3080 - beefy enough for this?

[Ryan Dunn @dunnry]

@Aaronontheweb @petabridge Gotcha. I am looking at maintaining a read index of all children of an actor. So, a Folder Actor that is parent to many File actors. I need a view that can quickly return a list of files. Was considering using this to update a FolderIndex actor containing list of all files.

[Ryan Dunn @dunnry]

Is there a reason to use Persistence.Query in https://t.co/ICz42WPypO now after Persistence added async callbacks to Persist(All)(Async) methods in 1.5.57? Can't you just update your read model there and not wait for journal (and deal with tagging)? @petabridge