@trq212 Can individual Claude instances communicate with each other?
E.g. as a solo dev I‘d like to have a Sales Claude that can check with Design Claude to create marketing material using the Figma MCP.
Separating instances in the hope of not diluting context.
@shloked@TownAI In general, is there a way to tune the writing style that you can recommend?
Simple few-shot prompting doesn’t do the trick for me, but setting up an Ax + GEPA pipeline feels like such an overload.. 😕
I’ve just released Byte Pulse, a handy little macOS app that displays the usage limits of AI coding tools in the menu bar.
Check it out here:
https://t.co/SFtPbTjC66
"Ein Lied über Reichtum im Stil von Sum41"
Das neue Audio-Modell Lyria passt in der @GeminiApp die Lyrics an den politischen Punk-Background an.
Bin beeindruckt!
Kimi-K2.5 ist für mich ein ziemlich spannendes Release.
Zum einen schließt es zu den SOTA-Modellen aus den USA auf, zum anderen lässt es sich auch ohne Quantisation auf einem realistischen System (2x M3 + 512GB RAM) ausführen.
Damit könnte man Tools wie @openclaw tatsächlich komplett lokal in guter Qualität ausführen.
Bin gespannt, wie es in den kommenden Tagen in der Coding-Praxis performen wird. Die Kimi-K2.5 SWE-bench Ergebnisse lassen viel Gutes versprechen!
https://t.co/AP7wUlMmPh
Beyond coding, it's hard to quantify the impact of new SOTA models like Gemini 3 Pro (https://t.co/xZmZouYFpI). Yet, my trust grows with every generation. I find myself fact-checking much less.
I think we're approaching a plateau for everyday use. Soon, the average user won't feel the model improvements anymore. The next big leaps will only show in specialized use cases, such as coding agents, scientific discoveries, etc.
Every start-up employee should be able to create their own (internal) apps for highly specialized use-cases, using Cursor / Claude Code / Lovable. This is especially true for small teams.
The productivity gains are simply massive.
Start-ups that do not actively encourage this will find it difficult to compete with young competitors in the very near future.
Without knowing too many details, I would also tend to favor an exit for >40x monthly profit, as sudden liquidity for a 6- to 7-figure asset is definitely a privilege.
Continuing to operate Supastarter puts you under pressure because, without subscriptions, you constantly have to find new users in your niche. In addition, maintaining always comes with opportunity costs.
An exit, on the other hand, gives you the opportunity for (temporary) financial independence. It also gives you the opportunity to set up your next project more professionally. In your case, for example, via a holding structure for tax advantages.
In the end, it's probably a trade-off between product (one-time payment –> pro exit) and momentum (vibe-coding hype –> contra exit).
@AntoineMinoux@TweetsOfSumit@isaacfrench_ +1, my girlfriend belongs to the Danish minority, and even with that background, it is difficult to obtain the right to own a Sommerhus.
It really is a strange situation in which properties across the border are about one-third as expensive as they are here in Flensburg. 😅
@AntoineMinoux@TweetsOfSumit@isaacfrench_ I regularly follow “Deutsche Grundstücksauktionen” for cheaper prices, but the main problem is that comparable properties to Isaac's sites in Texas (mostly in MeckPomm) are not approved for building.
USA: 38 inhabitants/km²
Germany: 237 inhabitants/km²
😕