Julien

@fuckthedip Suis sorti il y a peu

GTT Je n'ose rien dire pour éviter de me prendre la foudre de X Finances

@Finance__Lab @NotionRaphaxelo Oui ce sera bien pour mes stratégies daily stock picking, mais je teste des stratégies sur des ETN levier 3 sur le NASDAQ et les historiques sont courts d'un part, donc il faut faire un backcast à partir du sous jacent, mais si on remonte à 1999 : pas d'OHLC.

@cheulibrown @NotionRaphaxelo Tu as essayé de recup les données avec yfinance ? normalement ca marche bien sous python

Un taux de réussite de 72% peut cacher une catastrophe statistique. J'ai backtesté une stratégie "Contrariante" qui affiche 12 800 € de gains depuis 2010. Le chiffre semble propre, mais son Z-score est médiocre (0.82). Sans confiance statistique, ce n'est pas du trading, c'est du hasard pur. Arrête de copier des algos sur Twitter sans auditer la fiabilité des données. Je te donne la méthode pour scanner tes propres codes. Plus de détails dans ma newsletter de dimanche

@NotionRaphaxelo J'ai heurté le même type de limite avec Tradingview. Alors je bâtis les stratégies en python et je teste tout dans VS Code. La denrée rare maintenant je trouve sont les data en format OHLC en daily sur du très long terme.

@cheulibrown Si c'était possible avec PRT ça serait top :)

@7sAcHiN255 Not in my experience, unsuitable for film music, which requires precise control, making it possible to steer it in what the user wants regarding harmonic, rhythmic and textural content. It's not there yet, unless you're only churning out cartoons x.com/cheulibrown/st…

Gemini's Lyria 3 creations Can these advancing AI tools really be a harm for the creators in the music industry and filmmaking? The level of perfection these AI tools are having is insane. They can match a beat or render a scene with a precision that feels almost impossible.

@pwealthydad Unfortunately try prompting a specific film music track in the symphonic Hollywood style, it won't even respect harmonic content nor rhythmic instructions. Check my example: x.com/cheulibrown/st…

Most people will use Lyria 3 wrong 🤯 They’ll type: “Make a pop song.” If you want elite output, define the musical core: • Genre + era • Specific instruments • Tempo + energy • Mood + dynamics Layer details. AI music rewards specificity, not laziness. Precision in = quality out. Now add vocal direction and story. Don’t say “singer.” Say: • Male baritone or female soprano • Gravelly, breathy, soulful • Fast, laid-back, building You can even upload an image to guide the mood. Lyria reads emotion, setting, action. I’m dropping a copy-paste prompt template you can steal. [Genre/Era] track featuring [Specific Instruments]. The vibe is [Mood/Dynamics] with a [Tempo] tempo. Vocals: [Voice Type/Gender] vocals that are [Vocal Tone/Style]. Lyrics/Theme: A song about [Topic/Theme]. [Genre/Era] Mid to late 90s R&B slow jam track featuring [Specific Instruments] warm Rhodes chords, mellow sub bass, tight kick with soft snare, finger snaps, silky string pad, and a clean electric guitar riff. The vibe is [Mood/Dynamics] late-night, romantic, smooth, and intimate, with a gentle build into the chorus with a [Tempo] slow, swaying 70–80 BPM tempo. Vocals: [Voice Type/Gender] female mezzo-soprano vocals that are [Vocal Tone/Style] breathy, soulful, and controlled with tasteful runs. Lyrics/Theme: A song about [Topic/Theme] missing someone after midnight, replaying memories, and asking for one more chance. Style influence: Toni Braxton meets Brandy, with Boyz II Men-style harmonies for the hook. @GoogleAI @GeminiApp @GoogleDeepMind

Providing basic midi information as input to the model (notes, rhythmic patterns) would help steering it in the right direction, and then fine tune the production details with words. There's no AI orchestration tool yet available to musicians with such sound quality.

So it understood the terms of the assignment, but it fails to respect basic rules like "no rhythmic backbeat", "no rhythmic filler" that sort of ruins that kind of aesthetic. I understand it's just meant as a fun tool for now.

I've given Lyria 3 a pretty specific prompt of what makes a convincing chase/pursuit scene in the symphonic Hollywood trope. It turns out you can't steer the model, it just outputs what it can, not what you're asking for. @GoogleDeepMind

@roro76120 @fuckthedip Capital expenditures

@fuckthedip Question c'est quoi le CAPEX ?

#Amazon Est ce mal élevé si je dis concours de quéquette sur les capex entre les géants de la tech? Capex 2026 : ~200 Md$ ⚠️ nettement au-dessus des attentes (~146 Md$).

@demishassabis @GoogleAIStudio Is there any way you could fix the saving failure bug in Theme Park on 3DO ? It's so frustrating.

For example I’ve been doing a bunch of late night vibe coding with Gemini 3 in @GoogleAIStudio, and it’s so much fun! I recreated a testbed of my game Theme Park 🎢 that I programmed in the 90s in a matter of hours, down to letting players adjust the amount of salt on the chips! 🍟 (fans of the game will understand the reference 😀)

We’ve been intensely cooking Gemini 3 for a while now, and we’re so excited and proud to share the results with you all. Of course it tops the leaderboards, including @arena, HLE, GPQA etc, but beyond the benchmarks it’s been by far my favourite model to use for its style and depth, and what it can do to help with everyday tasks.

@NotionRaphaxelo Ta voix de stentor bien sûr ;) Tu automatises le passage de trades également ou c'est plus un outil pour traiter tes titres sous surveillance ?

@cheulibrown Merci du retour.. Qui est la plus forte des voix? (J'avoue que mes oreilles ne sont pas les organes les plus efficaces 😂) Pour l'UT je m'en sers en 4h sur un algo mais il est un peu modifié 😉

Un Dimanche soir. Je voulais tester un truc vite fait sur les Bollinger. 3 heures plus tard et 284 millions de trades backtestés, j’avais un algorithme capable d’optimiser les cycles de volatilité. Moralité : ne jamais “juste tester un truc” quand ton ordi s’appelle Divibot. Plus de détails dans ma dernière vidéo 👇

@fuckthedip In depuis avril 👍

Je suis en train de lire la publication de Wallix Elle a pris plus de 130% depuis le début de l'année, aujourd'hui ils "confirment" leurs objectifs, pas rehaussés, suffira t'il? Nous allons voir ...

@Eric_Lewin @marcfiorentino @bfmbusiness Et quand on est actionnaire Cogelec, ça se passe comment en pratique ?

Suspension de cotation pour Cogelec, l'une de mes grosses lignes chez @marcfiorentino sur @bfmbusiness

@Bloom__ow Regarde du côté des micro organismes "miroirs" (par la chiralité)

@barriere_dr @Bloom__ow Oui c'était de l'ordre de 13-15% en 2000. Cela privilégiait plus la capacité de travail brute, l'auto discipline que l'excellence scientifique par contre, je pense que la différence se situe surtout là.

@Bloom__ow Arf … 10% en 98 Ça va ? 😅

@karpathy Oh so that's what we call what I've been doing. I'm totally outside of your field and have been vibe coding trading scripts and backtesting them (which is also outside of my field). Some logic blocks eluded my comprehension at first but it teaches me. Copy-pasted errors as well.

There's a new kind of coding I call "vibe coding", where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. It's possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good. Also I just talk to Composer with SuperWhisper so I barely even touch the keyboard. I ask for the dumbest things like "decrease the padding on the sidebar by half" because I'm too lazy to find it. I "Accept All" always, I don't read the diffs anymore. When I get error messages I just copy paste them in with no comment, usually that fixes it. The code grows beyond my usual comprehension, I'd have to really read through it for a while. Sometimes the LLMs can't fix a bug so I just work around it or ask for random changes until it goes away. It's not too bad for throwaway weekend projects, but still quite amusing. I'm building a project or webapp, but it's not really coding - I just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works.