The Lettore

[The Lettore @TheLettore]

@cjav_dev @peppersilly I think it is not really clear: now the linked page says that, at the same time, the API credits cover the Claude Agent SDK and claude -p, but also that you can still use them with the subscription limits. How can both be used?

[The Lettore @TheLettore]

@cjav_dev @peppersilly And we can continue using claude -p with our subscription, can we?

[The Lettore @TheLettore]

In effetti però sull’app IO c’è questo avviso: “Puoi usare la tua patente su IO solo in Italia esclusivamente per dimostrare l'abilitazione alla guida in caso di controlli delle Forze dell'ordine”, che implica non abbia valore per altri usi (che è comunque un’idiozia).

[Luca Sofri @lucasofri]

Le risposte che ti dà il customer care AI di @Avis quando cerchi invano, nei giorni successivi, di farti rimborsare un noleggio auto che hanno rifiutato al banco pretendendo una patente fisica e non quella digitale ufficiale (sì, è uno di quei tweet sfigati di sfogo da truffati).

[The Lettore @TheLettore]

@pangram Are you working on an extension for Safari on iOS and/or an iOS app to which we can share text?

[The Lettore @TheLettore]

@bcherny @Hummer292 A thing that is not clear to me is why, when the cybersecurity classifier fires, the model gets downgraded to Opus 4.8 and not 5 (as with the biology one). I disabled the automatic downgrade and in these cases I switch to Opus 5 and it works fine, so why it is not the default?

[The Lettore @TheLettore]

@athyuttamre Please add proper support for turn-based translation. Unfortunately, the new Voice Mode keeps talking over the speaker, even when it’s explicitly instructed not to.

[The Lettore @TheLettore]

@athyuttamre @sammcallister It’s very good for everything but translation, in my opinion. Not because the translation ability is worse than the old Advanced Voice Mode, but because the new model doesn’t respect the precise instruction to wait for the speaker to finish talking, and it talks over him.

[The Lettore @TheLettore]

@RobertJBye Unfortunately it cannot work as a live translator in a bidirectional way, since it support only one language at a time.

[The Lettore @TheLettore]

@bcherny The biologic risk classifier is a little bit too aggressive, it blocks many of my perfectly safe clinical cases benchmark 😔

[The Lettore @TheLettore]

@_mohansolo Is this valid for Antigravity CLI too?

[The Lettore @TheLettore]

@btraut I switched from Claude Code to Codex recently and project-only hooks do no work :(

[The Lettore @TheLettore]

@noahzweben I hope the Claude Code Desktop app will support per project MCPs (as the CLI does), since without it a lot of my projects are broken on the desktop app (Claude can’t reach the necessary MCPs), and as a result I can’t use Dispatch.

[The Lettore @TheLettore]

@gabriel1 Isn’t it against WhatsApp ToS?

[The Lettore @TheLettore]

@superwhisper Any news about the ElevenLabs Scribe model in the Windows app? Been using it lately on macOS and it’s wonderful

[The Lettore @TheLettore]

@petergostev I think we already lost it with the new “channels” feature of CC

[The Lettore @TheLettore]

@amorriscode Thanks. Will it be possible to use project-level MCP without needing to use global ones, as it is possible with the CLI?

[The Lettore @TheLettore]

@RobertJBye Thanks, an update to the mobile app was sorely needed. Is there a specific changelog somewhere, or could you be a little bit more specific about the improvements, especially regarding large prompts performance, MCP connections, and attachment uploads?

[The Lettore @TheLettore]

@RobertJBye It would be helpful to include the section on usage limits, to avoid the need to visit the website to check it on mobile

[The Lettore @TheLettore]

@mikeyk Can we use our existing Claude subscription with it, in the same way as Claude Code?

[The Lettore @TheLettore]

@bcherny @0xRaduan I will try it, but I would like to understand if there are technical differences that would make it a better choice for non coding work, or if it is only a matter of UI and UX. Is there a different system prompt? Other technical differences that would warrant the switch?