mediablur

Exploring the intersection of AI & media tech | Focus on audio/video innovation | Passionate about empowering creative expression through emerging technologies

[mediablur @mediablur]

@sama Why did you roll out image generation to free users if you don't have the capacity and you are compromising the service for paying customers?

[mediablur @mediablur]

@testingcatalog No voice mode yet, but at least we’ve got something 👏

[mediablur @mediablur]

@altryne Hmmm. In testing with things I actually care about (rather than benchmarks), Grok 3 with thinking is giving me better results than Gemini 2.5 Pro. Choosing the right model for the job is still important.

[mediablur @mediablur]

@testingcatalog Heard any updates on when the Android app will be released?

[mediablur @mediablur]

Confused about what speech you hear is real and what is not? This won’t help 😉. Audio created using the new OpenAI voice model at https://t.co/cnpDr7x9Ux.

https://t.co/Z2JLAKLGtG

@OpenAIDevs @OpenAI

[mediablur @mediablur]

@cursor_ai Wow, these improvements are awesome! Thank you!

[mediablur @mediablur]

@AravSrinivas Would rather have the Sonar API support MCP.

[mediablur @mediablur]

@AravSrinivas YES! (please)

[mediablur @mediablur]

@kevinweil It would be great if you added API support for MCP to OpenAI models 🙏🏻

[mediablur @mediablur]

@AILeaksAndNews I'm hoping we get MCP, but of course that's Anthropic's idea so maybe not ....

[mediablur @mediablur]

@packyM That looks like the book I need right now. Just placed an order for a copy. Thanks.

[Linus ✦ Ekenstam @LinusEkenstam]

Look mom,

The Grok team change the system prompt to exclude:
M u s k & T r u m p

Sad to see them censor like this, do better.

[mediablur @mediablur]

@altryne Just a heads-up that Haize Labs is spelled incorrectly in the podcast notes for today's episode: "Verdict, a new open-source framework from Hayes Labs". I finally found the right company but thought I'd mention it in case it's something that can be fixed for others.

[mediablur @mediablur]

https://t.co/2guoKWFY1e

[mediablur @mediablur]

People in the AI industry assume people that aren't in the industry have the same level of understanding of the capabilities of the technology. Generally speaking, they don't. I think most people will be more confused than enlightened by the ad (if they noticed it to begin with).

[mediablur @mediablur]

These are the only models available on the current Android app with a Pro account:
* Default (Optimized for fast search by Perplexity)
* Claude 3.5 Sonnet (Most advanced model by Anthropic)
* Sonar (Advanced model trained by Perplexity)
* GPT-4o (Latest advanced model by OpenAI)
* Grok-2 (Latest model from xAI)

[mediablur @mediablur]

@altryne This is disappointing: o3 mini has a training cutoff date of September 2021? I tried to get it to list the discography of the group "The Armed" since the discography is not straightforward, but it could only come up with 2 albums and had the wrong title for both and the wrong date for one. Deepseek R1 did MUCH better.

[mediablur @mediablur]

@altryne @AITinkerers I missed hearing about this somehow (and I listen to ThursdAI every week). Is there a way to track when this event is in Denver/Colorado?

[mediablur @mediablur]

@AleRVG I showed the video to my wife and now she says that she wants a robot that is just like the one in the video. 🙂

[mediablur @mediablur]

@altryne That's awesome 😁