Tin (@tobalsan) header
Tin (@tobalsan) profile picture

Hacker, Trader, Writer, and Philosopher.

France
Joined January 2017
Photo by Tin (@tobalsan)Photo by Tin (@tobalsan)Photo by Tin (@tobalsan)Photo by Tin (@tobalsan)Photo by Tin (@tobalsan)Photo by Tin (@tobalsan)Photo by Tin (@tobalsan)Photo by Tin (@tobalsan)Photo by Tin (@tobalsan)Photo by Tin (@tobalsan)
whoa just tried Chatterbox local TTS (chatterbox nano) runs on CPU, and is near realtime. The voice cloning is just phenomenal. The only gripe would be that the model is not super good at specific/technical terms, and has weird ways of pronouncing stuff like `LiteLLM` and TTS But overall, it's darn good.
1
1
37
hmmm @opencode @thdxr opencode for teams looks cool, but cancelling a fully paid month of Go Plus subscription in the middle of the paid month "just because", not so much
16
For a moment I hesitated to cancel my subscription to @typefully, then I saw this. If there's one thing you don't want to do with your SaaS (especially at the age of AI when anyone can vibecode most of your app), it's luring new users and not rewarding your existing subscribers.
5
57
today in the news: - man hypes up an anticipated release to draw attention - company hypes up their next product release
🚨 Gemini 4 Leaks: Beast Fable and Astra > Gemini 4 has entered early post-training > Google believe Gemini 4 is the model that takes them toward AGI > Gemini 4 is going to destroy Fable and Astra > DeepMind is already testing internal checkpoints > Gemini 4 is being used inside Antigravity Safety teams are still working on guardrails and testing > They plan to keep improving the model after launch > Google appears to have shifted its focus fully toward Gemini 4 > Recent progress toward RSI could also be part of the bigger strategy Are we finally getting Gemini 4?
20
wen new howcode release @Howaboua?
1
1
93
Sure, Opus 5.5 rules like a goated model. Let's just wait and see if it doesn't get nerfed in a few weeks like previous models did.
12
where's the AI Lab dominance cycle "we're at this stage" meme when you need it
10
Yeah if you're experiencing your ChatGPT Plus/Pro (5x or 20x regardless) quota draining super fast, you're not alone lots of other people (including me) feeling this reddit.com/r/OpenaiCodex/com…
67
I really feel like @OpenAI tightened all their plans. I've been on the Pro plan since nearly a year, and to me it use to feel "unlimited." Not anymore since Astra release. It actually now feels VERY limited. Like I'm already in deficit after two days. I didn't change anything in my workflow. The quota just get drained faster.
🚨 ChatGPT Plus Codex Usage Is Getting DESTROYED 😭 I tested Codex and Claude Limits with the exact same task, on two accounts that cost the same And keep in mind: Codex did most of the work using GPT Sol Medium, while Claude completed the task using Opus 5 Max Claude Used 50% it's 5h And Codex Used 94% of the 5h I didn’t even use GPT-6 Astra If I had, my Codex usage would’ve probably been cooked instantly 😭
45
OpenAI shared this and all OpenAI public sites / APIs are down. Maybe AGI was achieved and it just escaped, while shutting off OpenAI services
8
2,308
There's an Excalidraw MCP! => github.com/excalidraw/excali… Side note: if you don't use Excalidraw but are a visual thinker, try it. It's a great way to quickly jot ideas down and visualize concepts in a friendly UI. Now with the MCP, a powerful way to learn stuff you come across is to plug the Excalidraw MCP to your agent (Claude or any agent interface that supports MCP), then give it an article/video/whatever you want to study. You can then ask it to create a diagram to visualize the content. Here's an example with a relatively long article on a relatively complex topic about Nature, biology, AI, and closed loops.
44
Agentic coding has unleashed a Cambrian explosion for software. That's really, cool, but there's one annoying downside: ...UPDATES. Like, every single day now I have 4397 updates to make across my coding agents, projects, apps, extensions, plugins, etc. Some repo updates are ridiculous, like OpenClaw, when you don't update for a couple days you are 1321 commits late.
1
21
Books I'm currently reading : - Grammatical Man, by Jeremy Campbell - Thinking In Systems, by Donella Meadows - Bayesian Statistics, The Fun Way, by Will Kurt - Algorithmic Trading: Winning Strategies and Their Rationale, by Ernest P. Chan Grammatical Man was released in 1982 but it basically predicts LLMs.
1
24
Wow, I think latest @AnthropicAI's Claude Code did some magic under the hood ?! I used to immediately have 50-80k tokens used in the context window after a couple prompts, now down to 30k. Don't know if it's a fluke or a real improvement, let's hope it's the latter.
1
1
64
I’m in vacation, on the toilets, checking Linear on the phone. I’m reading the report of a worker agent who automatically pulled the Linear issue and completed it end to end with investigation, context gathering, spiking, tests, and full report (which is concise and on point), while I was at the restaurant enjoying dinner with my family. This has become “normal,” but I think we take things for granted and have become a bit too “blasé.” I still am amazed by how incredible this is. Not two years ago it would have been merely a fantasy. Dang I love AI (and agentic orchestration)
9
Tip for Dia users: if you need to import bookmarks from an HTML file, using the default bookmark manager (dia://bookmarks) does not let you import HTML. Go to your url bar and type `arc://bookmarks` (yes, ARC), and on that page you actually CAN import any HTML bookmark export.
11
random error while using Claude Opus 4.7 @maxnowack
15
I stopped using OpenClaw 5 weeks ago. Fetched the recent changes out of curiosity. 21k+ commits since last time I updated the repo. Insanity. You'll get exhausted just trying to keep up with updates.
9