Opus 5.5 destroys GPT-6 Astra for marketing (7 real use cases)

Cloud Opus 5.5 versus GPT-6 Astra, which one actually wins, especially when it comes to doing actual work? So what you see on my screen right here is a comparison done by Artificial Analysis, and you can see, you can see the different, um, the different, uh, cost per intelligence and intelligence index. And you can see for this one up here, this graph over here, which is your, uh, the intelligence index versus Cost per intelligence index. You can see that GPT-6 Astra, let's just go to medium over here, seems to be a little lower when it comes to cost per task. You also can see the intelligence index that Claude Opus 5.5 is higher and the medium is higher over here. It's actually a little less cheaper. What I did for this experiment was we focused on medium because medium is a little more practical for real life. It's the same exact prompt and we just let it go. We just let it interpret things. And we just want to see how well it figures things out on its own. And keep in mind, this is using Medium. Not only that, we got to look at costs as well before I start to show you some of these examples. $4 per million input tokens and $20 per million output tokens. And then Astra lists at $10 and then $50. So it's significantly higher. And we're going to first start with the packaging part. So right now with the packaging piece, so this is this one over here. What the packaging looks like, and this is this these are the thumbnails. That's what it is. And so this is a thumbnail saying, "Oh, this is what you should fix first," right? So I'm going to show you the three thumbnails that it came up with. Here's another thumbnail on backlog debrief. So this one over here, show the show the source content draft approval block. Okay. So what do I think about these? I don't think these are great, if I'm being honest. Right? Like I don't first it's it does this whole this whole AI written thing where it kind of makes you feel like a like like a dummy, right? Like three promises, three ways in. Like what does that even mean? Fix this first. What does that mean? Backlog to brief. Like, what does that even mean? Show the source. What does that even mean? This doesn't quite make sense to me. So I'm curious to take a look at what Opus put together, right? So let's take a look at what Opus put together and then we can decide on a winner here. So if we look at this one, this is the tab where Opus did some work. So win the click over here. So it says 4 AEO mistakes. So is this the best thumbnail ever? No, but at least I'm in the picture, right? Not that I need to be in every picture, but At least I can understand where it's trying to go. Or calls should move to drafts over here. ContentOS will pull from your best content, whether it's internal meeting sales calls, and then it drafts from. It's at least explaining the logic to me a little bit. I guess it did a revision. Here's the original version and then it did a revision, but they look the same to me. What am I going to say about this? I'm going to say this is a lot more usable. Again, they both had the same exact prompt. It seems that Opus 55 had a little more horsepower here. So you can see, you know, it even kind of explains its reasoning as well over here. I'm like, hey, that's cool, right? So in this case, I have to give the first round to Opus 55. And keep in mind, Opus 55 medium and then GPT-6 Astra medium. That's what we're looking at. So round 2, let's go ahead and look at round 2. So round 2, again, I'm going to explain it real quick. So round 2 is an ad teardown. Okay. We're back on ChatGPT over here, and you can see that what we're the the directive here is inspect the supply ad and then rebuild around an offer the business can actually fulfill, and then competitor performance is unknown. So what is this over here? So let's take a look at the ads. Okay, so ad number one is this. Okay, Signal Foundry. Okay, all right. So the platform for every content job. So this is talking. This is just like a fictional ad, right? This says planning, drafting, refreshes, reporting, one workflow, many possibilities. So This, we have a workflow for this, which does AEO and SEO. I'm like, well, I don't quite understand that, right? And then you have one offer, you have 3 different test creatives. Okay, your backlog isn't a priority list. So again, this is just AI slop, if I'm being honest with you. So my interpretation so far based off these first 2 examples is it looks like the horsepower from a thinking standpoint is a little bit lower when it comes to GPT-6 Astra Medium. So I'm going to move over to the second one and we can take a look at second one being Claude. So let's take a look at that one. And you can see for Claude, here are the ads that it drew out. These are the teardowns, right? And so you can see this one, 60 deals in flight, zero time to type. Talk your way through admin, works in Salesforce, Gmail, and PandaDoc. Bad prompts equals bad results from ChatGPT. Speak rich. So this is like, I think this is pulling from Whisper Flow. That's what it is, right? And so this one over here, meeting notes that get the deals, the details right. I think this one, this one actually came up with good ad concepts for Whisperflow, and I think it just decided to pick something versus just, I don't know, for, for chat, it kind of just didn't really come up with a concept to go after. And I guess maybe it's pulling from one of the workflows that we have. So I look at this and I'm like, okay, look, there's more bad prompts equal bad results from ChatGPT. AI skips your brand. Here's one that got you cited. Okay. So basically, maybe it's looking at Whisper Flow over here. It's looking at what's working and it's saying, hey, here's what you should go after over here for single grain AI scripts or brand. Here's one that got you cited. 400+ more AI citations for Ellis Building products. So yeah, that's actually pretty good. I understand what it's doing now. It's actually saying, hey, based on these Whisper Flow ads, these are what we think you should go after. And then here's 3 options that you should go for, which I think are right here. So solid SEO missing in AI search. Okay, 400+ citations. Read the case study. This is not a bad ad. This is pretty good. And we're not even using the full power of Opus 5.5 here. So this one over here, AI skips your brand. Here's one that got you cited. Now, do I like this italic? I don't like these italicized fonts now because it screams that it's slop, but from AI, like AI design slop. Here's what we built, stair framing, roofing, and decking calculators, rewritten product pages. I get it. It's trying to cite a case study that we did. I see where it's going, and this is where you have to work with it a little more because even when I do my YouTube packaging, for example, with my packaging skill, I still spend 5 to 10 minutes on it. So that human taste and judgment layer that seems so cliché does matter quite a bit. And so again, I'm going to say for round 2, in terms of rescuing the ad, Opus 5.5 is the winner. So we are 2 for 2 right now. We are now on number 3. So number 3 is a lead journey investigation. So this is a synthetic nested cohort where people leave and page evidence suggests what to test. It does not establish a cause, right? It's looking at a funnel. It's looking at 800 people land on a landing page, 320 actually click the call to action, 80 form start, submit, book. This is like a fictional cohort over here of 800 users. So it's then recommending what to fix. So here's a promise mismatch, 7 asked, 3 needed, no clear next step. So it's saying that the offer is not good. It doesn't match the promise. It's saying the form has too many form fields. And then 3, the handoff, there's no clear next step. One page, one next move. So it's saying before it was all this stuff over here, and then all these questions, and then it's saying, here's what's— here's What's proposed. That kind of makes sense to me, right? And then we can go over here, 7 asks, which we kind of talked about already. No clear next step over here. So it's saying proposed measurement is this one. So the promise mismatch, the hero offers a free audit, the business offers the 20-minute triage call. So we need to make sure that we align that. So these aren't necessarily bad. I would say these are fairly decent suggestions. If you're enjoying this video right now, you gotta check out Single Grain. Single Grain is where we have our AI-native marketing team Helping you with AEO, SEO, helping with paid media, helping you with creative scaling, all the things that you need to grow faster to get more customers in a world where you just need more distribution. Go to singlegrain with a G, singlegrain.com, and we'll see you on the other side. And by the way, like I've been very big on GPT-6 Astra for a while now, but here's round 3, finding the leak. Okay, so this is interactive, which is cool. So it shows stage 1 is a blog post. Okay, stage 2 is a call to action click. So you have 10. Post captured, 4 different button labels, and then lead form has 4 forms and success message over here. So what is it saying exactly? So it's saying kind of a similar thing. Form A asks for 8 required inputs before a first conversation. So it's saying too many fields, too much friction. The success message is a dead end. And then 3, the call to action promises free marketing plan, the form never mentions one. So this is actually talking about the single grain website. So let's just look at the fixed one prototype. So the before was accelerate your marketing without headcount. After is about you, and then it'll route your request after. So you have to fill these out first. And so the required inputs before was like 8, but I like how it breaks it out a little bit. You know, it's actually 3 plus 2 moving forward, right? About you is over here, right? So 1, 2, 3, and then route your request would be 2 more. So it kind of splits up the form field a little bit. Measurement plan, turning friction into a measured result. Okay, here's what it says. All the event names are over here. Form completion rate, guardrail. The success message is a dead one, right? So get your free Free marketing plan. Before was accelerate your marketing without getting headcount. So it fixes this one over here. And then this one is perform never mentions. UX is a little more intuitive than what I was seeing. And I think the proposed fixes are a lot more, you can kind of digest them easier, right? And so I'm going to say this is a perfect UX? No. But do I understand the changes it's asking me to make? Yes, I do. So I think this is I think again, I would give the edge here to Opus 5-5. So 3 for 3 right now. So let's go ahead and take a look at the next one. Number 4 is stall deals. So explain the silence, like why aren't deals responding to you? And then show the next move. Okay. So let's take a look at this one. This is number 4. So explaining the stall. So what happened exactly? Why are we not winning these deals if we've talked to them already? Like why are they ghosting exactly? And this is again, all fictional data. So it's saying, hey, Maya Chen spoke to Rowan from this company, but we don't know the replacement. This person doesn't seem to be the right fit. So basically what we do know, we asked for a fee scope breakdown, the contact, and then whether this unresolved question is the main reason for silence. Okay, so the owner is Maya over here, Maya Chen. Verify the responsible contact first. Once confirmed, send the original scope and review map with context, not imply the new person agreed to pilot. This is a draft that it's created. So it's like, hey, this other contact, we had talked about a proposed 30-day content pilot. Before we take it further, I'd like to confirm whether you're the right person to review scope. Okay? So the proposal was this, I can share the scope, or should I work with somebody else? Okay. So it's basically saying, okay, this is not the right person. And it's trying to reach out to somebody else to try to keep the deal moving. And that's often the most important thing when it comes to sales. The fortune is in the follow-up and most people just aren't good at following up. And so this will help fix that, right? Or this one over here, hey, we discussed this over here and this seems to be the main thing. So D3 explicitly requires product team approval of technical claims. So yeah, your team would approve technical claims. So it's just reaching out to people for technical review friction or budget is unavailable. Your team would approve technical claims. It seems to be the same thing. So I'm not a fan of this UX because again, I'm getting confused again here. I don't know what this means. The buyer asked, the follow-up missed. What does that even mean? Well, Let's move on to Opus 5.5 and let's take a look at what the output looks like. Opus 5.5 here. So explaining the stall. So closed loss of September 18th, no reason recorded. So if we take a look at this one, I have to stare down a little bit in the camera. So CRM closed the deal because the field is empty, confirms the stall. Let's just take a look at some of these screens over here. So last buyer message was this. Last seller follow-up was this. Loss reasons. Okay, we're not filling this stuff out. Okay, I get it. Fine. Let's take a look at this timeline. Okay, here's a timeline from the seller to buyer, the back and forth over here, which is, you know, I suppose helpful to look at. And then also you can take a postmortem. Okay, what E10 does and doesn't tell us. Okay, so it stalled after here. Here's the hypothesis: the highest— the price was way above what they planned to spend. Proposal grew past what they came for. Decision sat with someone else we never met. Okay, here, and then here's what it says. Okay, what the close tells us: the approver is still unidentified at close, and then proposal could have tested E8 through E10, ask who else weighs in and by when, offer a 20-minute walkthrough for anyone absent from the calls. So this just gives you a sense of what it would do. And this is the first build over here. Momentum stopped at the proposal, scope landed, price was higher than expected. Next move is this over here. Then it did a revision, which I like. And then it shows all these numbers down here. So I think at least for me here, chronologically and from a design standpoint, I understand what I need to do next. Whereas the other one, I get confused. And confusion is a nice way way to just get nothing done. So am I going to say either of these are really good examples? No. But which one would I prefer? Opus 5.5. So far, Opus 5.5 is 4 of 4 versus GPT-6 Astra. Okay, so, so far it's 4 for 4. And if we're taking a look at round 5, this is case study and proposal. So make the proof and the scope easy to see. Okay, so let's take a look at this one, client-ready deck. So this is what a deck would look like. And you can see, okay, This is Juniper's past demo result and BeaconOps proposed work. Stay separate, navigate with the arrow keys, the deck and scope are downloadable. So a clear queue, a common review. We can create drafts, we get stuck in deciding what to do and who can approve it. A focused 30-day content operations pilot. So let's take a look at this. Give every draft a clear owner, inventory sources, product editor, review time unchanged, causality is unproven. This is a pilot. Here's the baseline. Small batches, visible review. Okay. A scope you can point to. 30 days with dependents approved and scope. Okay. It's clearly AI slop, but at least I understand what it's trying to scope out. This was going to require more work. I'll just kind of leave it at that. And then let's take a look at what Opus 555 put together. So here it is. And by the way, they even gave me a deck PDF over here if I want to use that. So we can see here, okay, paid search isn't producing qualified pipeline and AI answers barely mention Prospect B. So $129K paid search spend, 15 MQLs in the last 90 days, $20 cost per qualified opportunity, and then 9.23% ChatGPT share of voice versus 80% of competitors 1 and 2. Okay. I at least understand, I better understand the problem set here in terms of like what's not going well. And then for this one, it says client A had a thin site. Single Grain built pages and tools around one path of quotes. So this is like a case study over here, I think. Before, built, and then— I think you should probably talk about examples here, like case studies, but numbers, but maybe it's over here. Yeah, this is the past results. So yeah, it talks about the case study over here, which I just mentioned. So that's good. Proposed 4 work streams. Okay, that's good. Scope is only what Prospect B selects, terms, values, conditions. Okay, great. And then 5 phases gated by decisions, no date terms. Okay. Total fee is TBD until 4 decisions are made. I like that because we're not pigeonholing— pigeonholing ourselves, right? So I would say that both of these decks require more work, but I like to think of Toby Lutke from Shopify likes to call these slop grenades where you throw these AI-generated pieces of content that frankly haven't been reviewed and require more work on the human side. So whatever you can do to reduce cognitive load for the other people that you're working with, the better, right? So I would just say that if you are reducing cognitive load, that actually means that you have more intelligence with what you're putting into something, more horsepower. And so 5 for 5 so far in terms of something becoming client ready. The next one's actually really exciting to me because we're talking about editing something, editing a video, and we're going to see, and this is going to be my first time looking at it, we're going to see which one's actually better. Let's take a look. By the way, if you need a connector that combines all of your tools into one area where your agents can talk to you, we have our Single Brain Gateway. Think of it as an API gateway where you can have all your connectors, you can define what your goals are, and then we'll even float the right workflows you can use from our side that we use for our clients, such as our cold outbound infrastructure or ad creative scaling. These things you can actually just use yourself. Once we decide, or once you decide what your goals are and you put in your connectors, then we can float the right workflows over to you. But the Single Brain Gateway is completely free to get started. That's where you can have all your connectors. You do need a connector if you don't have these already. That way you can start to build other cool things such as a single brain or having memory, shared context, all these things that you need to have your agents running well. singlebrain.com/gateway and we'll see you on the other side. So this is the next one and this is the the edit. Let's take a look at this edit and see how good it is. Okay, so this is a. I basically had it. I took an entire video and I said, hey, pick the best spot that you like and edit it to your liking. And to be frank, this does look like AI slop, but we'll see how it does. Okay, so here we go. Grokbot will help you conquer AEO and SEO. So on my screen, we have an AEO SEO bot over here, and you can see what it does is it pulls from content that we create. It also pulls from internal conversations that we're having, whether it's customer calls or prospect calls, and it gives you content spikes, content ideas to go after. And you can see over here, it gives you a bunch of ideas and you review if you want But people often forget that AEO SEO is still very much a real game and you do need to publish content. And you can see it'll define the shape of the problem, the funnel, and if it's net new, is it an update, is it consolidation? And so what happens here is then you can draft more content using the ClickFlow MCP and API. You can connect other different tools like Google Analytics, Google Search Console. And then what's going to happen is it's going to allow you to skip. Okay. I don't think it's particularly— I mean, you can clearly tell it's AI generated and there's no reason to have some of this other stuff down here. But I like the fact that it's a split screen and you can see me doing the screen share. So it's pretty clear and there's captions. And so this is passable. And to be frank, I've been testing videos like this right now and some of them have been working. So I'm actually very curious to see. This could be my first time looking at this. Let's see what Opus 5.5 put together. So let's take a look. So it took the original video here and it's showing me the before and the after, which I actually appreciate. Actually, to be fair, the ChatGPT one does show that as well. So am I going to say either of these are really good? No. And actually what's interesting to me is I've seen other people increase the thinking level and their edits are actually pretty good. There's actually more going on. I would add more B-roll. I would add examples like screenshots, things like that. But I like the fact that ChatGPT chose a moment where I was doing screencasts and I think that one's actually better suited because it's actually more interactive and there's more stuff going on. This one, I think it's a good moment, but We need better edits for this one overall. And yeah, that's what I would say. So I actually think for this one, from what I've seen just with this example, I would say that ChatGPT-6 Astra or GPT-6 Astra actually wins that one. Okay, so now last but not least, we're on to our final one. So let's take a look at our final one over here. So the final one basically is around We are talking about AEO engine. So we're talking about turning a content backlog into a clear next move. Okay. So that's the next one. So this is talking about a clear next move for every page. Okay. So inventory, this is your content inventory over here. So tofu, mofu, bofu, all the content over here. How many clicks you got in this period over here? Here's all the tofu content. Here's all the mofu content. I get it. And then diagnose, okay, the traffic is wrong. Decide, what do you actually decide to go after? What do you decide to draft? What do you decide to review? So this is like a, I think it's a workflow for deciding, like a content operating system for deciding what actually gets published and what doesn't. So we have something that's more advanced than this, significantly more advanced where you don't have to do all this work already, but I just wanted to see what it'd come up with. So that's the example from ChatGPT. And I just, again, there's too much cognitive load here. When I'm looking at this one, I'm like, I kind of don't know what to do. I'm just gonna— if I don't know what to do, I'm gonna get frustrated. I'm just gonna move on, right? And so let's take a look at what Opus 5.5 put together. So it says over here, okay, so here's the dashboard, right? 3 evidence-linked recommendations. So you should be updating this piece over here, consolidating this one, new page over here. These are— and by the way, it tells you what to consolidate, what to update, what's in that new page over here. And then these are all the pages, right? So, and then if I click through, Let's just look at what we are approving over here. It shows a draft. Do we check the claims on what we're making over here? Are these claims actually verified in terms of what we're publishing? We should be checking that. Here's the before and after for the revised draft. That's cool. So it just seems like there's more going on. It's easy to understand, right? What do we update? What do we consolidate? What's a new page? What do we keep exactly? This is ready for handoff over here to a human being. This at least more thought has gone into this UX overall, and you can see over here. Oh, this is pricing is actually missing over here. Here's what you should do about it. And so obviously now with tools like Jev, you can do this a lot faster. You just need some type of workflow. Sometimes maybe you might need a UX, but you know for us, you know most of the work are most of our workflows are done with the agents, and then you know we just kind of review the work afterwards. Again, I would just say overall when you look at all the work that's been done. The only one that really won very slightly was the edit. And I truly think that had we put Opus 5.5 on high, it probably would've performed better. But again, we're talking about practical, more frequent use cases, and you're probably going to set it to medium. But 6 of the 7 rounds that we ran over here with the showdown, Opus 5.5 was the winner. And I didn't even know until I showed you guys. Oh. Anyway, your mileage may vary, and I would suggest that you test both. And in fact, I pay for both, so I might as well just use both. But Anyway, hope you enjoyed this video and we'll catch you in the next one.