Master negotiators? Or do they know something we don't?
open.substack.com/pub/handyai/...
15.12.2025 14:28
👍 0
🔁 0
💬 0
📌 0
Master negotiators? Or do they know something we don't?
open.substack.com/pub/handyai/...
GPT-5.2 is here and, while it’s cool to see an incremental release like this have such wild leaps on benches, I’m starting to get more concerned about corps rushing through safety checks to outpace each other.
Yesterday, Suno announced a $250M raise and Udio announced (another) settlement, this time with Warner Music Group. Why the divergence?
open.substack.com/pub/soundmac...
Google‘s new image model is their Sora moment. Nano Banana Pro seemingly allows any copyrighted content and the image outputs are scary realistic with incredible details around fidelity (up to 4K!) and crisp text generation.
Sora 2’s separate feed solves some problems, but the model itself (namely the Cameo feature, which allows you to insert real people into AI videos) is going to create some new, hard-to-solve problems. More:
handyai.substack.com/p/sora-2-and...
Sora 2 today
Suno is teasing its next generative music model (v5). Their short comings have been vocal quality and general fidelity and this sample seems to indicate some solid progress on both. Take a listen:
Poke is a marvel. The bot went from insulting me, to gaslighting me, to haggling with me over setting a price for its services. In the end, I got access to the entire service for free.
Here’s how it went down.
open.substack.com/pub/handyai/...
Modern product manager must be moving beyond simple wireframes and into building full, functional prototypes with AI coding tools. No excuses.
Read how: open.substack.com/pub/handyai/...
Interesting tidbit: ChatGPT’s study mode defaults to a 10th grade level. Does OpenAI consider this the statistical middle in terms of Chat users?
🚨🚨🚨🚨
The spread of intelligence between the various GPT-5 models is pretty wild.
In response to user complaints last week, ChatGPT now has three modes for users:
- Auto
- Fast
- Thinking
Always use the Thinking mode. OpenAI increased the limits from 100/wk to 3,000/wk so there’s no real reason not to, and this is the version of the model that’s topping all the benchmarks.
technically not available yet! sometime today
YouTube link is live for GPT-5 livestream (starting in an hour-ish)
www.youtube.com/watch?v=0Uu_...
Grok 2 and 3 will be open sourced... any day now... I'm sure it's coming...
GPT-5 tomorrow
it appears as though ElevenLabs has created an Udio-quality competitor with licensed training data and some pretty solid commercial terms
more: soundmachines.substack.com/p/the-first-...
ditch your accountant, use Cursor's AI agents
open.substack.com/pub/handyai/...
o3 still the king of cost-per-intelligence, generally. Good showing by Grok 4 but it was overhyped (who expected anything different?)
DON'T: Ditch #AI tools entirely. You’ll be at a detriment.
DO: Set limits and be mindful about usage. Protect your brain!
More: open.substack.com/pub/handyai/...
Welcome to the age of misalignment.
www.anthropic.com/research/age...
thanks Music Ally for featuring me here. lots to chew on in terms of how hackathons are going to look in the future
musically.com/2025/06/13/v...
