Corentin

664 posts

Corentin banner
Corentin

Corentin

@corxntyn

Passionately curious. Web agency COO 🇨🇭 Building: https://t.co/tcneaRtXCb

Prague Katılım Ekim 2022
152 Takip Edilen104 Takipçiler
Corentin
Corentin@corxntyn·
The thing I miss most in Belgium is fries. So I can't really follow a diet when visiting 😤
Corentin tweet mediaCorentin tweet media
English
0
0
1
48
Corentin
Corentin@corxntyn·
@full_kelly_ @elonmusk Bro literally owns SpaceX which has a strategic partnership with Cursor that also includes formal option to acquire them for $60B late 2026
English
0
0
6
594
Elon Musk
Elon Musk@elonmusk·
Try Composer 2.5
BridgeMind@bridgemindai

New CursorBench results just dropped. Two big takeaways. Composer 2.5 is way better than most people think. 63.2% score at $0.55 per task. Nearly matching Opus 4.7 Max and GPT 5.5 Extra High at 20x less cost. This is insane value. Gemini 3.5 Flash is #10 at 49.8%. Below GPT 5.5 Low. Below Opus 4.7 Low. Google's newest model can't even beat budget tier competition. Composer 2.5 is the sleeper. Gemini 3.5 Flash is the disappointment.

English
2K
3.3K
21.2K
8M
Thomas Paul Mann
Thomas Paul Mann@thomaspaulmann·
Gemini 3.5 Flash is now available in @raycast: 🧑‍🏫 Better at following instructions ⚡ 4x faster than previous models 🖼️ Sharper multi-modal reasoning Enjoy!
Thomas Paul Mann tweet media
English
10
7
150
12K
Corentin
Corentin@corxntyn·
@peduarte when can we expect new models like Gemini 3.5 Flash to be available in Raycast AI chat? 👀
English
0
0
0
262
Corentin
Corentin@corxntyn·
@peduarte really enjoying it so far, especially dictation
English
1
0
1
385
Pedro Duarte
Pedro Duarte@peduarte·
how are you liking the new raycast so far?
English
118
0
185
21.4K
Corentin
Corentin@corxntyn·
/goal is one of those small features that changes the whole agentic experience. I love it. Must use in Codex and Claude Code. You stop writing tiny babysitting prompts like "now run tests, fix this error, do this, check this again" etc. Instead just tell it what "done" means. "/goal make this work properly, run the tests, fix what breaks, and don't stop until it's clean"
English
0
0
0
99
Corentin
Corentin@corxntyn·
Crazy how fast things change in AI. A few months ago I almost cancelled my ChatGPT account and mostly switched to Claude for work. Now I'm back to ChatGPT and Codex because it's crazy smart and just works really well.
English
0
0
1
83
Corentin
Corentin@corxntyn·
@huntie If you're on Mac looks like the public beta is out!
English
0
0
1
85
Corentin
Corentin@corxntyn·
After 24 hours of using dictation in Raycast, I can confidently say they cooked and nailed it. No need for another app like WisprFlow or a separate subscription anymore. It's insanely fast and accurate.
Corentin@corxntyn

Another thing I'm really excited about in @raycast v2 is speech-to-text dictation🎙️ Tested it briefly this morning and it's fast. WisprFlow-level fast, even in terms of accuracy. Very good start. One thing I'd love to see in upcoming updates: more privacy controls. Auto-delete history and transcripts after X days, clear local data, etc. Maybe also the ability to use your own API keys and choose the STT model you want. Dictation is insanely useful especially when working with Codex/Claude Code. I've tested so many: WisprFlow, MacWhisper, Spokenly, and VoiceInk, which I've been using lately. But having it directly inside Raycast is awesome.

English
2
0
15
4.6K
Corentin
Corentin@corxntyn·
@myPhilTaylor Yep, looks like it's using Scribe v2. But I'm happy dictation is built right into Raycast now. Hopefully there'll also be a way to use BYOK.
English
0
0
1
126
Corentin
Corentin@corxntyn·
Yeah, I don't think AI is anywhere close to doing that reliably yet. There's still too much hallucination and risk. But we're heading there eventually I guess. Also, frontier models that handle tool-calling well are still very token-hungry, expensive and need huge amounts of compute...
English
0
0
0
24
Thiago Salvador
Thiago Salvador@bettercallsalva·
@ileppane @corxntyn @raycast Two things slow this. Reliability under noise (mic picks up off-topic and the agent dispatches wrong tool), plus the trust gap of letting voice trigger destructive actions. Saw a colleague almost ship a Slack DM to wrong channel via voice agent.
English
1
0
0
45
Corentin
Corentin@corxntyn·
Another thing I'm really excited about in @raycast v2 is speech-to-text dictation🎙️ Tested it briefly this morning and it's fast. WisprFlow-level fast, even in terms of accuracy. Very good start. One thing I'd love to see in upcoming updates: more privacy controls. Auto-delete history and transcripts after X days, clear local data, etc. Maybe also the ability to use your own API keys and choose the STT model you want. Dictation is insanely useful especially when working with Codex/Claude Code. I've tested so many: WisprFlow, MacWhisper, Spokenly, and VoiceInk, which I've been using lately. But having it directly inside Raycast is awesome.
Corentin tweet media
English
7
0
34
3.1K
Corentin
Corentin@corxntyn·
Honestly I can't tell a difference. So far it's been decent, fast, and accurate. You can also have your own dictionary, instructions + style for each app. I was also using WisprFlow before switching to VoiceInk with BYOK because I didn't want to pay a monthly sub for dictation. Love that I can have everything in one app now.
English
0
0
1
191
Brian Coords 💻
Brian Coords 💻@briancoords·
Can I just run a normal command in Claude Code - like.. instead of asking it, I just type the command and it runs it? Or do I always have to have a separate tab open? It'd be nice to just not burn tokens interpreting a command but also keep it in the conversational context.
English
9
0
2
1.7K

Keşfet

@full_kelly_ @elonmusk @thomaspaulmann @raycast @peduarte @farzyness @huntie @BarackObama