Simon

Content Creator || Romper Enthusiast || Loophole King || 100k+ TikTok #homingfromwork

[Simon @SimonPeacocks]

New deepseek via opencode is buzzzzzzzin 🔥

[Simon @SimonPeacocks]

5 years in and the best we’re doing is *checks notes* doing dinner reservations? They’ve got balls releasing this lmao

[Simon @SimonPeacocks]

@satyanadella Drill me is crazy work

[Simon @SimonPeacocks]

As much as i like the t3code app, imma need voice integration to actually consider it in my workflow. Might be🍴time

[Simon @SimonPeacocks]

So I’m about 30m away from a nice little vibe station. Then comes the hard/fun part - reading!

[Simon @SimonPeacocks]

I basically setup a corpus of German translations and tagged them with translator decisions. Then, I’m running my target book through the tagger and we automatically query the dataset for similar prose across the other books

[Simon @SimonPeacocks]

So in the end, I get to read along with context on the German across other books and make decisions according to my own vibe AND with the context of how other people have dealt with similar writing. There are a ton of opinions on how to translate - so in the end, it’s vibe imo

[Simon @SimonPeacocks]

@thsottiaux Ah oh.. and stability in the mobile apps. Both apps start freezing after like 5-6 running chats and loading times are abhorrent. With respect :)

[Simon @SimonPeacocks]

@thsottiaux Adding live voice to the mobile apps would be nice - these days I’m always remoting in.. so much so that i haven’t even gotten to use that cool feature yet

[Simon @SimonPeacocks]

@jiang_kev @polarbrowser Using agent-browser from vercel.. Beats most things I've seen. Anyone compared this yet?

[Simon @SimonPeacocks]

It’s a wild book! Quite fun, I’m enjoying this process

[Simon @SimonPeacocks]

Has anyone worked on book translating using models? I’ve been working on an en of “Das Reich Ohne Raum” 1925

It’s slow going.. one interesting thing has been using bertalign + kimik3 to compare multiple translations of existing works and understand the decisions made

[mal @mal_shaik]

claude code opus core

[Simon @SimonPeacocks]

@bentlegen See though I prefer this to under zealous opus who I have to poke every turn to even try. easy enough to handycap solo, I can't to shi to motivate a claude

[Simon @SimonPeacocks]

@itsgreyum @bentlegen Codex doesn't even allow that natively. Must be using an alt harness which, still surprising behavior

[Simon @SimonPeacocks]

opus 5 is so bad its genuinely a liability. but, i still have a week left of tokens😖 wish there was some sort of token resale market. hard to imagine a model this bad can be released by a major lab in the current year.

[Simon @SimonPeacocks]

@JinaAI_ is the /v1/train api down? I tried training requests on v3-5 embedding models and I'm getting a 500 error

[Simon @SimonPeacocks]

@ctatedev I just ran this by an agent and it informed me of this thing exactly. Y'all are cookin 🤌🏾

[dax @thdxr]

used a trick @jlongster came up with

agents can control browsers but you can also ask it to record network requests into a HAR file

then it can derive a client for any website which is more efficient than browser controlling it every time

made it build a quick uber eats cli

[Simon @SimonPeacocks]

@SuCoffey123 @VictorTaelin It's valuable, yes.

Proof is that programming was supposed to be dead 1-2 years ago and the biggest AI companies are still buying out open-source for the talent.

Knowing computer shi puts you in a better position with got tier models. Godspeed