Valdas

[Valdas @valdasio]

@thdxr yes

[Valdas @valdasio]

@0xSero Even codex and claude code on mobile is lacking, I think the best experience is tailscale with termius

[Valdas @valdasio]

@Teknium A tiny local model can achieve similar results and could also be useful for the smart approvals mode. It's still early, but have you seen these posts:
https://t.co/ENd0afVW7c
https://t.co/MqndIZsMyc
https://t.co/7A7Ir9IQ2i

[Theodore @theoleecj]

On the Jev waitlist?

You can use not Jev (but close enough) right here: https://t.co/S0EQnHkSGL

https://t.co/qLtRs0fHyD browser demo

[Valdas @valdasio]

@thdxr I think local is the way, there's already a bunch of DIY implementations. For example https://t.co/kiD9Ol7VOu

[Harsha Gundala @harshagundal]

They were building in stealth for 2 years, I was building in stealth for 2 hours…

Happy to open source Qwen-2.5-1B-RLCD, 5x faster on-device inference for JSON workloads that need to be type-safe.

⚡️Demo below on a M4 MacBook⚡️

every LLM has the ability to efficiently batch inference every key of a JSON at the same time and generate probabilities from a set of possible categories. No new training required, but it’s easy to optimize if you need!

On hugging face now!

[Valdas @valdasio]

@shinboson Nice work! What is it running on?

[Valdas @valdasio]

@opencode Do we know the context size or vision capabilities?

[Valdas @valdasio]

@ItsmeAjayKV Impressive. How does it compare to your other tested models?

[Valdas @valdasio]

App 26.908.70816 (9275), macOS 26.6.2

[Valdas @valdasio]

Anyone else seeing high usage with Codex desktop when idle?

My M4 mac gets hot with no tasks running. After ~4.5 hours open, Codex Renderer was using ~5.5 GB memory. During a quick diagnostic chat, Renderer showed ~27–31% CPU and Codex Service another ~30–35%. @thsottiaux

[Valdas @valdasio]

@ashen_one Opencode Go currently gives you $60 worth of usage on the $10 plan

[Valdas @valdasio]

@mattyp Loving the Grok Bot Galaxy stream so far.

One thing though - people keep having to re-explain context that another team already has.

Any plans for a way to share a team so bots can actually collaborate across accounts?

[Valdas @valdasio]

@bridgebench Any ETA for DeepSeek-V4.1-Flash scores?

[Valdas @valdasio]

@mattyp @poteto @roshan_s @bot But will Grok 4.7 release in time for that 🤔

[Valdas @valdasio]

@bridgemindai Any bets for grok 4.7?

[Valdas @valdasio]

@bridgemindai Seems to be the case for everyone

[Valdas @valdasio]

@dhh Does it also support the GPU?

[Valdas @valdasio]

@omkarships Easy - add a paywall 😎

[Valdas @valdasio]

@kelvinbuildss Not without distribution, for instance, a personal brand. Without an existing audience, you'd likely need some budget to gain initial traction and make sales.

[Valdas @valdasio]

@simonecanciello hmm, what about instead of measuring weight, you measure consistency? Seems like the retention would be a lot better and would build a stronger habit 🤔

[Valdas @valdasio]

@Samaytwt Postgres is the most popular for a reason