The Fork in AI: GPT 5.5 Unveiled  🔗 **Source:** [The Fork in AI: GPT 5.5 Unveiled](https://share.snipd.com/episode/a291c10a-0c83-4d7e-b0d4-70efbc1d6617) ## Highlights - **Apple Lets Siri Run Third Party AI Models** - Apple will let users pick third-party AI models for Siri via an extensions system in iOS 27, iPadOS 27, and macOS 27. - Users can assign models like Gemini, ChatGPT, or Claude and even set custom voices per model, changing Siri from a single-stack assistant to a pluggable gateway. Transcript: Jaeden Schafer They are finally cracking Siri open. Mark Gurman, he first was reporting on this whole framework back in March, but there is a new report that came from 925 Mac that is confirming that this is actually going to ship at WWDC On June 8. This is going to be part of iOS 27 and iPad OS 27 and Mac OS 27. So all of the latest updates are gonna be coming out in June. And the mechanism is basically called extensions. What this means is that you're going to be able to go into settings in Apple intelligence and Siri, and they're going to pick what model handles what so you can plug Gemini in you can plug Chat GPT and you can plug Claude in and what's interesting is you can actually set custom voices depending on which model is responding. Enough feature to apologize to everyone for why this has all taken so long, especially at the same moment that this lawsuit has come in where basically they're going to have to pay $100 Back to everyone that bought an iPhone 15 Pro and an iPhone 16 because Apple intelligence didn't ship when those were launched and the marketing all basically pointed to it shipping. ([Time 0:00:55](https://share.snipd.com/snip/eaac346c-f92f-4e94-98ee-06b500fb0ad5)) - **Chrome Downloaded A 4GB Gemini Model Without Consent** - Google Chrome silently downloaded a 4GB Gemini Nano model into a folder called optguide without prompting users, raising privacy and storage concerns. - Researcher Alexander Hanf verified the install from macOS kernel logs and argued the behavior may violate EU e-privacy and GDPR rules against storing data without consent. Transcript: Jaeden Schafer There was a security researcher named Alexander Hanf. He's a computer scientist and a lawyer based in the EU. And he caught Google Chrome downloading a 4GB Gemini Nano model onto the user device. There was no consent prompt. There was like, it didn't ask you. It just downloads this 4GB model onto your computer. The file is called weights.bin and it lives in a folder called optguide on device model. Apparently, according to Hanoff, he verified the install by reading macOS kernel file system logs. There's a bunch of reasons why this is problematic. According to this particular researcher, Hanoff, he says that this violates the e-privacy directive article 5.3, which prohibits storing data on a user's equipment without prior Consent, plus GDPR Article 5 and 25. So, you know, this isn't really super great to give four gigabytes onto, you know, a billion plus Chrome users computers without any sort of, you know, telling them that they could do This. And the precedent isn't great if, you know, every AI company could start downloading models onto your computer and say it's just, you know, part of their software upgrade. I mean, these models, the nano ([Time 0:02:01](https://share.snipd.com/snip/57c1e49c-a65e-4a53-9697-b7b1df2f322b)) - **Host Promotes AIbox.ai To Combine 80 Models For $8.99** - Jaeden Schafer described his product AIbox.ai as a consolidation platform offering 80 AI models for $8.99/month to avoid multiple subscriptions. - He explained practical use cases like using Claude for writing, ChatGPT for images, and 11 Labs for audio without separate accounts. Transcript: Jaeden Schafer And speaking of multiple different AI models, if you're already paying for ChatGPT and CloudPro and Gemini and Grok and maybe 11 Labs and all of these different audio image AI tools Out there, I'd love to tell you about my software company called AI box.ai. We have one platform with 80 different AI models on there, everything from the top AI labs, and it is $8.99 a month. So you get access to all of the top image, audio, video, music models in one place for one subscription. You don't have to have tons of different accounts. And most importantly, you don't have to have $20 subscriptions on 20 different accounts. They cost you a ton of money. So hope this saves you a lot of money. It is AIbox.ai. It's how I recommend people get access to all of the different AI models, use them all, test them all, try them all. I love using Claude for writing and tons of different tasks, but it doesn't have an image generator. So I got to go over to chatgbt to do that. And then that doesn't have, you know, an audio MP3 file generator. So I go over to 11 labs for that. Anyways, all of those in one platform, um, you can go check it out links in the description. ([Time 0:03:13](https://share.snipd.com/snip/11de4573-915c-4102-a44b-600e0a1301eb)) - **GPT 5.5 Instant Cuts Hallucinations On High Stakes Tasks** - OpenAI released GPT 5.5 Instant as the new default across free, Plus, Pro, and Enterprise, claiming 52.5% fewer hallucinations on high-stakes prompts. - Benchmarks improved (AIME math 65→81, GPQA 78→85, ChartVIV 75→81), targeting law, finance, and medicine reductions in errors. Transcript: Jaeden Schafer First of all, they have a new model GPT 5.5 instant. This is going to be the new default on CHAP GPT. This is rolling out to everyone on free plus pro enterprise, and it basically replaces GPT 5.3 instant. Maxwell Zenith did a whole write up about this and basically said that the headline number on this is that opening I keeps telling everyone is that there is 52.5% fewer hallucinations. That's the claim that opening eyes making on this. And these like the lower amount of hallucinations are specifically on tasks that they are deeming as high stake prompts. So that's medicine, law, and finance. They also say that there is 37% fewer inaccurate claims on tough conversations that had, you know, previously been flagged by users. So they were able to kind of look at what was flagged. They trained and made this better. According to the AIME math, it jumped from 65 to 81. GPQA went from 78 to 85. CharVIV went from 75 to 81. So basically a bunch of benchmarks. We moved up. ([Time 0:04:17](https://share.snipd.com/snip/78ec9b67-7790-483d-adb7-53a48086f5bd)) - **Check And Edit ChatGPT Memory Sources** - Use ChatGPT's new memory sources panel to inspect, delete, or correct which past chats or connected data (like Gmail) informed a personalized answer. - This prevents mistaken personalization, for example ChatGPT treating a friend's car as your Toyota Corolla because of prior shared context. Transcript: Jaeden Schafer They also said that there is a memory sources panel. So when ChatGPT personalizes an answer using any of your past chats or any of your kind of saved memory or your Gmail or any of that kind of data, you can now see which memory it pulled from And you can also delete or correct it. Now, this is actually something that I really appreciate because I've had many conversations where, you know, I'll have a friend that like use chat GPT or my wife will use chat GPT for Something. And now when I'm talking about it, it's like grabbing the context of that conversation. And it's like, since you're, you know, I don't know, worried about this, your Toyota Corolla, blah, blah, blah. And I'm like, oh, man, like I let my friend ask a question to chat GPT or ask a question to chat GPT about his car. And now it's, you know, thinks this is my car forever, right? So there's a whole bunch of cases where this can happen. You can delete that. And I think that's super useful. ([Time 0:05:33](https://share.snipd.com/snip/4c3dbd47-3880-4149-9669-78608f180c59)) - **Try OpenAI Self-Serve Ads For Premium Audiences** - Consider testing OpenAI's self-serve ads manager for premium customer acquisition because bids start at recommended $3–$5 per click and it's now open to all US advertisers. - Expect tools like conversion API and pixel tracking; OpenAI aims for $2.5B ad revenue this year and targets high-value verticals like legal and medical. Transcript: Jaeden Schafer And so in line with all of that, OpenAI just flipped the switch on self-serve ads manager for all US advertisers. So it is live right now. You can go to ads.openingai Personally, I think this is probably an incredible opportunity to get your products in front of people that previously, you know, I mean, this is a very niche Audience. It's cost per click bidding. There's recommended starting bids of $3 to $5 a click. So OpenAI is really targeting, you know, premium, more premium products, $5 a click. I mean, lawyers, dentists, doctors, these people can afford it. If you're selling a hot tub, yeah, you can afford it. But there's a lot of cheaper items that, you know, t-shirts from Timu are probably not going to be making any money on these $5 AI clicks. There is a conversions API, there's a pixel measurement, and there's kind of a whole performance marketer toolkit. They dropped this. And they used to have a $50,000 minimum spend. So that was kind of for like a pilot that they were running, they don't have that either anymore. And according to Sarah Fisher at Axios opening eyes, internal goal is to get $2.5 billion in ad revenue this year and $100 billion by 2030 from this self-serve ad program. ([Time 0:07:01](https://share.snipd.com/snip/66a9aec5-34c7-4340-acba-6ed8b20eaa02)) - **OpenAI Plans Roughly $50B Compute Spend This Year** - Greg Brockman testified that OpenAI will spend about $50 billion on compute this year, a jump from about $30 million in 2017, signaling massive infrastructure demand. - That spend flows to hardware vendors (e.g., Samsung) and helps explain surging chip and memory market profits tied to AI growth. Transcript: Jaeden Schafer Greg Brockman just finished his testimony in the Musk-Altman trial over in Oakland, over on CNBC. They were talking about this. They were in the room. Apparently, Greg Brockman confirmed under oath that his personal stake in OpenAI is now worth about $30 billion. And also what I think is interesting, he said that OpenAI is going to spend about $50 billion on compute this year. That is, I mean, just compute. That is wild. The number was about $30 million in 2017, and now we're going to $50 billion in 2026. It was tens of billions last year,, and I think 50 this year, that is a huge jump. I think they also previously disclosed a $600 billion spend pledge through 2030. And they said over a trillion and change in basically all of their spending has been announced, but we didn't know exactly what was happening this year. So that's, that's pretty big. Also, as it just as a side note, this is happening at the exact same time that Samsung just crossed a trillion dollar market cap today. And this was all basically on the back of AI memory demand. And they did an 8x profit. The profit went up 8x because of this. ([Time 0:09:10](https://share.snipd.com/snip/194c87ef-bab9-4bcd-ad5f-ed9c4674d258))