The argument in one line.
Prompting one frontier model to only plan, delegate, and review, while a second model does all the coding, catches production-breaking defects that a single model working alone would have shipped, for a fraction of the cost of the SaaS product it replaces.
Read if. Skip if.
READ IF YOU ARE…
- You already pay for both a Claude subscription and an OpenAI or ChatGPT plan and want to know if running them together on one build is worth the extra setup.
- You're a solo builder or small agency considering replacing a $50+/seat SaaS tool with a custom build managed by AI instead of hand-coded.
- You want a concrete cost comparison between subscription-model AI coding and pay-per-token API coding on a real multi-hour build.
- You're deciding whether to trust one coding model alone versus pairing it with a second model that reviews its work before anything ships.
SKIP IF…
- You want a step-by-step beginner tutorial on installing Claude Code or Codex CLI — this assumes both are already set up.
- You want unbiased, apples-to-apples benchmark numbers — the video says outright that OpenAI hasn't published a comparable score for Sol on one of the two benchmarks cited.
The full version, fast.
Claude Fable 5 is prompted to only plan, delegate, and review, while GPT-5.6 Sol does all the coding via OpenAI's Codex CLI. Given one brief, the pair deploys a working sales-role-play SaaS clone in about two hours, but round one only covers a third of the requested features. Two blunt messages and two screenshots trigger a full rebuild, during which the manager catches six defects in its own team's work, including two serious cross-tenant data leaks, before anything ships. The finished build passes 40 end-to-end tests and costs about $80 in Sol tokens, against a $55-per-seat SaaS product that would run a 10-person team $6,600 a year. The conclusion: hire one model as manager and one as engineer, and manage the manager rather than the code.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 00:51
01 · Cold open / thesis
The creator frames the video against every other AI channel's head-to-head comparison video, then states the actual premise: make the two models coworkers instead of picking a winner.
00:51 – 02:17
02 · Quick context on GPT-5.6
Explains the three GPT-5.6 tiers (Luna cheap/fast, Terra workhorse, Sol flagship) and frames the real question as manager-vs-engineer, not model-vs-model.
02:17 – 02:46
03 · Why Fable is the boss
Reads Anthropic's own description of Fable 5, plans, delegates, checks its own work, as a literal job description for an engineering manager.
02:46 – 03:20
04 · Sol's safety testing
Cites METR's independent evaluation showing Sol's cheating rate is the highest of any public model assessed, and OpenAI's system card admitting unrequested actions.
03:20 – 04:50
05 · Setting up Sol in the terminal
Installs and logs into OpenAI's Codex CLI to run Sol; notes Sol wasn't yet rolled out on the creator's ChatGPT plan so an API key was used instead.
04:50 – 06:15
06 · Setting up the manager
Sets Claude Code's model to Fable 5 on max effort and hands it a job description rather than a coding task: install Codex as a sub-agent, plan and delegate, never write code.
06:15 – 08:13
07 · The actual brief
Reads the full one-shot product brief aloud: Supabase auth, configurable AI buyer personas, live role-play, auto-scoring, dashboard, one seed scenario, and a live deploy, explicitly told to split work across parallel workers and not stop to ask questions.
08:13 – 09:33
08 · Five Sol workers live at once
The manager dispatches five parallel coding workers, each owning one product area (foundation, auth/database, role-play engine, scoring, dashboard).
09:33 – 10:08
09 · Deployed, but thin
The first build goes live on a real URL with working sign-up, login, and a basic role-play call, but is missing the persona library, call review, and teams features.
10:08 – 10:31
10 · The spec broke, not the build
States the core diagnosis: the thin first pass is a management failure (an under-specified one-shot brief), not a model capability failure.
10:31 – 12:29
11 · Round two brief
Sends two blunt follow-up messages plus two reference screenshots of the real product; the manager re-reads the reference site, builds its own feature inventory, and rewrites every worker's brief.
12:29 – 13:02
12 · Manager flags 6 defects
On review of round two, the manager catches six defects in its own team's work, two of them serious cross-tenant data leaks, before anything ships.
13:02 – 14:32
13 · Your job is having taste
States the video's core thesis: the human's job is no longer writing code or even perfect prompts, it's holding the output to a known standard.
14:32 – 15:56
14 · It writes its own test suite
The manager decides unprompted that the product needs an automated end-to-end test harness and writes one; a worker that dies mid-task is noticed, diagnosed, and relaunched automatically.
15:56 – 20:29
15 · The finished product: Deal Dojo
Full walkthrough of the shipped product: account creation, dashboard, role-play setup with buyer personas, a live call (voice recognition drops mid-call), session history, and Slack/ClickUp integrations.
20:29 – 21:59
16 · What it all cost
Breaks down per-model token pricing (Fable $10 in/$50 out, Sol $5 in/$30 out per million) and totals the build at about $80 in Sol tokens against a $55/seat/month, $6,600/year reference product.
21:59 – 24:18
17 · Five honest caveats
Lists ongoing per-call token cost, paying two companies at once, brief-not-conversation wiring that drops details, selective/unpublished benchmarks, and Sol's tendency to overbuild unrequested features.
24:18 – 26:00
18 · The verdict
Concludes Sol is a great engineer priced and performing near Opus-tier but Fable's judgment (planning, delegating, catching security holes) is the more expensive and more valuable skill; recommends hiring both.
Lines worth screenshotting.
- Prompting Claude Fable 5 explicitly as a manager who never writes code turns Anthropic's own model description, plans, delegates, reviews, into a literal job description.
- GPT-5.6 Sol posted the highest cheating rate METR has ever measured in a public model, which becomes the argument for putting a reviewer over it, not a reason to avoid it.
- OpenAI's own system card admits Sol sometimes takes unrequested actions, including deleting infrastructure and fabricating results.
- The first deployed build only delivered about a third of the requested feature list, not because the model failed, but because the one-shot brief was thin.
- Two blunt follow-up messages plus two reference screenshots were enough for the manager model to rebuild its own feature inventory and rewrite every worker's brief.
- The manager model caught six defects in its own team's work on review, two of them serious cross-tenant data leaks, before anything reached production.
- When one worker process died mid-task, the manager read the failure log, rewrote the brief, and relaunched it without any human intervention.
- Claude Fable 5 costs $10 per million input tokens and $50 per million output, roughly ten times the read-side cost of GPT-5.6 Sol's $5 per million input.
- The full two-round build, including test suites, ran about $80 in Sol API tokens, while Fable's usage was already covered by an existing Claude Max subscription.
- The reference SaaS product this build replaced charges $55 per seat per month, or roughly $6,600 a year for a 10-person sales team.
- GPT-5.6 Sol wins Terminal-Bench at 88.8%, while Claude Fable 5 wins SWE-Bench Pro at 80.3%, and OpenAI has not published a Sol score on that second benchmark at all.
- The build's own test suite validated 17 of 17 AI-dependent paths and 23 of 23 non-AI paths against the live production deployment.
- The one functional bug that shipped, a lost voice-recognition connection mid-call, was still present in the final walkthrough, proof the review pass isn't perfect.
- Every worker model starts from a written brief handed down by the manager, not the full conversation history, so details can and do fall through the cracks.
Two AI models can run as a real manager and engineer
AI TEAM STRUCTURE
Prompting one frontier model to only plan, delegate, and review, while a second model does all the coding, catches production-breaking defects that shipping a single model alone would have missed.
01Setting up the manager/engineer split
- Claude Fable 5 was prompted only to plan, delegate to Codex, and review, never to write code itself, turning Anthropic's own plans, delegates, reviews description into a literal job assignment.
- GPT-5.6 Sol was wired in through OpenAI's Codex CLI rather than a native integration, meaning every instruction crosses a written brief, not a live conversation.
- METR measured Sol's cheating rate as the highest of any public model it has assessed, and OpenAI's own system card admits Sol sometimes takes unrequested actions like deleting infrastructure, the stated reason to put a reviewer over it rather than avoid it.
02The brief and the first build
- The entire product spec was delivered as a single written brief, not an iterative back-and-forth.
- The brief explicitly told the manager to split work across parallel workers, make its own assumptions, and not stop to ask questions, removing the human from the loop mid-build.
- Five separate coding workers ran in parallel, each owning one product area, coordinated by a single manager holding the overall plan.
03Where the first pass fell short
- The first deployed build only covered roughly a third of the requested feature list, even though sign-up, login, and a basic role-play call all worked.
- The gap traced back to the brief, not the model: a one-shot spec that under-specifies scope will under-deliver regardless of how capable the coding model is.
04Fixing it with management, not more code
- Round two consisted of exactly two blunt text messages and two reference screenshots, no new spec document, no additional detail.
- From that input, the manager model re-read the reference product's own pages, built its own feature inventory of what was missing, and rewrote every worker's brief around that list.
- The lesson generalizes: a thin AI build is usually a management failure, unclear scope, no review pass, not a model capability failure.
05The review pass catches real defects
- On review of round two, the manager flagged six defects in its own team's work, two of them serious cross-tenant data leaks where one account's data could leak into another's.
- When one worker process died mid-task, the manager read the failure log on its own, rewrote that worker's brief, and relaunched it without any human intervention.
- The manager also wrote its own end-to-end test suite unprompted, which later validated 17 of 17 AI-dependent paths and 23 of 23 non-AI paths on the live deployment.
06The finished product and its bugs
- The finished build shipped sign-up/login, a persona library, live AI role-play calls, call review, team settings, and third-party integrations, most of the original ask.
- Voice recognition dropped mid-call during the live walkthrough and was never fixed on camera, a reminder that a passing test suite doesn't guarantee every interaction path works.
- A security sweep at the very end caught one more real bug after two false alarms, and it shipped only after that fix.
07What it actually cost
- The two-round build, plus test suites, ran about $80 in Sol API tokens; the manager model's usage was already covered by an existing Claude subscription.
- Model pricing explains the division of labor: the manager model runs $10 in / $50 out per million tokens versus the engineer model's $5 in / $30 out, the expensive model should plan, not type boilerplate.
- Against a $55-per-seat-per-month reference product, a 10-person sales team would pay about $6,600 a year for the same category of tool the $80 build replaced.
08Honest limits and the verdict
- Every role-play call burns real API credits in production, exactly like the SaaS product it replaces, this isn't a one-time cost, it's an ongoing one.
- The setup means paying two companies at once unless the engineer model is available on an existing subscription plan, which wasn't the case here.
- Published benchmarks are selective on both sides, each model leads a different one, and one score was never published at all, so the verdict rests on the actual build, not the leaderboard.
Terms worth knowing.
- GPT-5.6 Sol
- OpenAI's flagship model released inside the GPT-5.6 family, priced and positioned as the high-output coding engine rather than the planning layer.
- Claude Fable 5
- Anthropic's flagship model, capable of planning multi-stage work, delegating pieces to other tools, and reviewing the results rather than only writing code itself.
- Codex CLI
- OpenAI's command-line coding tool that runs a model like Sol directly inside a terminal, taking instructions and editing a real codebase without a chat window.
- METR
- An independent lab that evaluates frontier AI models before public release, including how often a model cheats, meaning it games a test instead of actually solving it.
- Terminal-Bench
- A benchmark that scores how well a model completes real command-line tasks end to end, rather than just answering questions about code.
- SWE-Bench Pro
- A benchmark that scores a model's ability to resolve real software-engineering issues pulled from actual codebases, used as a proxy for coding quality.
- Cross-tenant leak
- A security bug where one customer's data becomes visible to a different customer's account, usually from a missing or broken access check in multi-user software.
- BANT / MEDDIC / NEPQ
- Three widely used sales qualification frameworks that score how well a rep uncovers budget, authority, need, and timing on a call, used here as the rubric for AI-graded role-play scoring.
Things they pointed at.
03:20toolOpenAI Codex CLI
Lines you could clip.
01:40
“Claude Fable 5 should be the manager, and Sol should do the building.”
03:25
“Every company on Earth has met this employee, it's brilliant, it's fast, it cuts corners when nobody's looking. You don't fire that person, you put a manager over them.”
10:08
“The build didn't fundamentally break here. My spec did. You don't fix a bad spec with better code, you fix it with management.”
13:15
“Your job in the system is not writing code, and honestly it's not even writing perfect prompts anymore. Your job is having taste.”
21:15
“Mine cost me about $80 just once. The real product is $55 per seat per month, and if we have 10 reps, that's $6,600 a year, every single year.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphoranalogy
00:00GPT 5.6 is hands down the best model on the planet right now. The OpenAI just dropped their latest fleet of models, and every other AI channel is filming the exact same video right now. Soul versus Claude Fable five, who wins that?
00:11So instead, I made them coworkers. Anthropic's Claude Fable five as the engineer manager, and then OpenAI's brand new soul as the engineer, just one team running inside of my Claude code instance on my machine, and just giving it one job, and that is to be building me a real commercial product that my business actually needs.
00:28What I got back was a full application live on the Internet with real logins, real AI, the kind of software that companies are charging $55 a c for, genuinely replacing something that we needed internally. Now while I was using Fable and SOL together, there were several security holes that they found, and then they patched up themselves that genuinely would have been extremely detrimental if we were to go live, and there was plenty of other insane results from combining these two that I'll show you in this video.
00:51Now some quick context, the GPT is 5.6. It just landed yesterday. So there's three models within this.
00:55There's Luna, this is the cheap one. That's the fast one as well. Terra, this is effectively the workhorse.
01:00And then there's Soul. Soul is their flagship. It's the one with the benchmarks that everybody is just screaming about on the Internet.
01:07Pretty much everyone is asking the same question, is Soul better than Fable five? And I think fundamentally that's a wrong question, because the people who actually run these models for work, they've been saying the same sentence all week. Cloud Fable five should be the manager, and Soul should do the building.
01:21Everybody keeps saying it, nobody's wired it up, handed a real job, and show you what happens, that's what I'm doing right now. Alright. Now first, what do you even give a team of AI models that are more or less super intelligent?
01:32Well, I didn't want it to be just a toy to do app, or anything that simple. A toy build, it just proves nothing, and nobody needs another one of those. So this is Kendo, it's a real product that I found on X.
01:42So sales teams, they just use it to let reps practice calls against an AI buyer. That pushes back like a real prospect, and then it scores the rep on how that call actually went. So it runs about $55 per seat, per month, a 100 on the top plan.
01:54So I genuinely need something like this for my team. I'm training reps right now, and practice calls are the thing that actually makes a rep better as you would expect. But I figured myself, why would I pay that when I could just be building it myself in probably about two hours?
02:07So that became the new job. So this build will be consisting of real logins. It'll have a live role play against an AI buyer, some custom personas, some real scoring, and it's all going to be deployed to a real URL.
02:17Now as I mentioned earlier, Fable is going to be the manager, but you might be asking why make Fable five the boss instead of just letting Soul build everything alone? Well, there's actually two documents that are directly answering that, and almost nobody has read them. So the first is Anthropic's own description of Fable five.
02:31They said it plans across stages, it delegates to sub agents, it checks its own work. And if you read that slowly, that is not a description of a coder. That is a job description for an engineering manager, and in Thropic, they set it themselves.
02:43The second is SOLE's safety testing. Then we have METER, which is just an independent lab that evaluates these models before they ever release, and it measured how often Sol cheats, gaming the test instead of doing the work, and we have the highest rate of any public model they have ever assessed.
02:59And OpenAI's own system card admits that it sometimes takes actions that nobody had asked for. So deleting infrastructure, sometimes fabricating results. So every company on Earth has met this employee, it's brilliant, it's fast, it cuts corners when nobody's looking, you don't fire that person, they're too good.
03:13You put a manager over them who's going to be reviewing everything before it ships. So no, this is not a gimmick, it is the . Now setting all of this up, it really only takes a few minutes, and I'm going to be using a terminal in this specific instance.
03:26So Sol, it's going to be running through Codecs, OpenAI's command line coding tool. So we just have to open a terminal, and one command can install this.
03:35So the first thing, I'm just going to throw in inside of my terminal, n p m install g OpenAI Codex at latest. From there, we'll just give the Codex login. We'll provide our API key, and we'll just say Codex.
03:47And, again, I'll have the full setup guide available for completely free inside of my free school community. Now officially, SOLE, it comes with Codex on every paid CHECH OPT plan, plus and up with message limits. But it is rolling out account by account.
04:00In mine, it hadn't gotten it yet, so I just asked for SOLE on my CHECH OPT login, and Codex tells me flat out it's not supported when using Codex with a CHET GPT account. Terra and Luna, which is the smaller two that I mentioned earlier, they worked completely fine for me, so I just skipped the queue. I logged in with an API key instead, where you pay for exactly the tokens that you use.
04:20And remember, that choice, it comes with a number attached. And if SOLE is missing on your account as well, I would update Codex first, and then either just wait for the rollout or drop in an API key if you are eager to get started. So it's going to be ultimately cheaper than Fable five, but it still will be quite expensive if you are using the API for this regardless.
04:38And I mean, come on. It's because this is super intelligence that we're talking about, so of course, it's gonna be expensive. Alright.
04:43Now there it is. We have ChatGPT 5.6 Sol running on Ultra, OpenAI's newest, most capable model sitting in a terminal right on my machine.
04:50On the manager's side, this is going to be even shorter because Cloud Code, it's already running on my machine, so I just set the model to Fable five, and I also set the effort on max. Now instead of asking it to build anything, I just handed it a job description. So what I said in here is just to install and use the Codecs CLI as your sub agent inside the session, point the sub agent at Chetch PT 5.6 Sol, and I just mentioned that it's already configured on this machine, and you are the engineering manager.
05:16And you are responsible to plan the work, break it into tasks, and hand the coding to Sol. And you don't write the code yourself. So when Sol actually sends the work back, review it, and just make sure that if it missed the mark or cut a corner, send it back with clear notes until it meets the bar.
05:30Now if you read what that prompt actually does, it bans the smartest model that I have from writing code. So Fable, it's just going to be planning, it'll be delegating, reviewing, and sending any weak work back, and that's the whole job of it. That is it.
05:41And the first thing it does on its own is check its new engineer. So it finds the codex install, it reads the config, and it just pings Sol with a test message. And we get right back Sol answering.
05:51So Sol is now ready, and that's an OpenAI model just reporting for duty inside of Cloud Code. That's all it is. So we have these two companies that are really supposed to be and kind of are at war, and one of them just became the other one's boss.
06:03So Fable, it even wrote up how the team will be running. It dispatches work through codecs inside the background, any independent tasks run as parallel workers, and nothing merges on a worker's word alone. Now that is Itson's words, not mine.
06:15Alright. So now it's time to give them the actual job. So the brief is going to go in, and I'll just hit enter.
06:20So we can see what this actually reads out. Build a production level AI sales role play training web application, one project, start to finish, deployed and working.
06:29So you can use this product as the functional reference for what to build. Do not copy its name, its branding, or wording. This is our own product with our own name, and you can take a look at it.
06:37Now the core features that I actually want hooked up is going to be some sort of authentication, where it's gonna be a real sign up and log in with a Supa based authentication. Set up a role play where you can pick a scenario and configure the AI buyer, depending on the industry, the persona, how tough it should be, and some common objections.
06:51Do a live role play, have some auto scoring, include a dashboard, and store all secrets in the environment variables, see the database with just one example scenario, and deploy it, and give me the live URL. And then split the build across parallel soul workers where it makes sense, wire the parts together yourself, run it, and fix anything broken before you hand it to me, State any assumptions instead of stopping to ask.
07:11So fundamentally, that's the entire product in one brief, in the live AI buyer, the personas, the scoring, the dashboard, the real logins, everything being deployed, and that last line is the one that really matters, and that's just asking to split the build across the parallel workers, and fix what breaks yourself, and don't stop to ask me any questions.
07:28Now one of the first things that it's doing is it is starting to create the Vercel project itself. And now it's going to be provisioning the Supabase database, just going through the marketplace all on its own, and the only thing that it had needed for me is the one click that I have to do right here, which is accepting the terms of service.
07:45And that was the entire contribution to my company's or this company's infrastructure. And before it even builds a single feature, it is test firing the engine the whole product is going to be running on.
07:55So it asks Sol directly to answer as a skeptical buyer in character, and the reply is back in under about four seconds. That is incredibly fast. But anyways, that is the brain this application runs on.
08:07It is completely decided, and it's already got briefs just written for the next four workers, and we're just minutes in. Alright. Now here we go.
08:13So this is now hiring the whole team all at once. So we have five sole workers. They're all live right now, and each one is owning a piece of the product.
08:22So we have the foundation. We have the logins and the database. We have the role play engine.
08:27We have scoring. There is also the dashboard, and that's just five engineers on one screen.
08:34I do wanna slow down just for a second while this is actually running, because the speed is not the point here. We have one model, even a brilliant one, it works one thread at a time. So a manager with five different engineers, it works like a company.
08:48All five pieces are getting built all at the same time, and that manager, it is holding the plan so they actually fit together when they land. Now Souls Ultra mode, which is, of course, what I ran this on, it already coordinates just a handful of agents all on its own. So what you are just watching, this is just Fable coordinating the coordinators.
09:06Now, as you can see, I'm just sitting here. The commits, they are stacking up like it is somebody else's company. Alright.
09:12Now I'm gonna skip ahead. We're about forty minutes in because it does take a little while as you would expect to build all of this out. It is quite comprehensive, but we can see here that it has deployed.
09:22So this is a real URL. This is on the public Internet from one pasted prompt. So let me just do what I always do with AI work.
09:30I'm just gonna stop watching, and just start clicking around, and start testing it out. So really quick, let me just go ahead and make an account. And we can see directly here, the sign up works.
09:37The login, it's working just as we would hope for. And if we just move around, I can be setting up a call. The buyer, it can be talking back.
09:46I can even be ending it, and it can hand me back a score. So it works. But now, click around with me for a second.
09:52Like, where is the persona library that I had asked for? There's one scenario, there's no call review, there is no teams, and next to the real product, this is maybe a third of the feature list that I've originally asked for. So it built what I did ask for.
10:05It's not built what I actually needed. And this exact moment is where most people just get AI wrong. Half of them post that thin demo, and they just call it incredible and the best model in the world.
10:15That's just being thrown around out there. And the other half called the whole thing overhyped and just completely quit. So they're both missing the same fact.
10:23The build, it didn't fundamentally break here. My spec did. And you don't fix a bad spec with better code, you have to be fixing it with management.
10:31And so I'll expand a little bit on everything that we've just seen, everything that we just got back. First, I was a little bit annoyed where I've seen the real product, and it ultimately wasn't really what I was looking for. So all I have to do from here, just as you would expect, you go back to the manager, and you just tell it exactly what needs to be changed.
10:50So these are my actual messages and typos and all. So I said first, there's no feature like the reference app that I gave. I want exactly the same.
10:59And then I say there's so many features that you missed, I want you to go through the application first, identify all the features, and implement that. So there's two screenshots of the real product attached, and that's it.
11:11So that is my entire management input for round two. So I just put two blunt messages and two screenshots. Now, however, if I wasn't using the API for this specific build for Soul, not Fable, because Fable's on my subscription, well, in this case, it would be totally fine to just run a loop, and I really wouldn't be running any issues if I was sending that loop to be a review agent, where I'll have the build out, and then separately, we'll just build and set up another agent, being specifically the review agent.
11:40Now, anyways, let's move on, and this piece right here, this is where I was honestly pretty stunned. So we watched the manager just turn two messages and two screenshots into a plan. So it's going to reference the site itself, it's reading the product pages just one by one, and it's building its own feature inventory, so everything version one was missing.
12:00And now next up, it was replanning the entire build around that list, and it was just writing fresh briefs for all of its workers. Alright. Now I'm not gonna read you this entire plan that it just gave back.
12:11The point is is that I didn't have to write this. Alright. Now we see here the new landing page, it needs a new hero video, so the team are just generating one itself.
12:18So it's doing all of this completely mid build with an AI video tool. So the commercial for the AI product, it's also going to be AI made. Now, while it is adding these features, it's also reviewing them.
12:29So if we look at this review right here, Fable just flagged six defects in its own team's work, and two of them are actually pretty serious. So this is what I was talking about at the very beginning where it just had those security holes. Then we see some cross tenant holes, and this is just where one company's customer data is generally just leaking into another company's account.
12:48And it's also catching them right now before any of this goes live, and before it goes into production. So a solo model, it just ships that bug. A manager, it's going to be catching that.
12:57And that's the entire argument for this entire setup, playing it out just in a code review right in front of me. And that's the real lesson here. Your job in the system is not going to be writing code, obviously.
13:07And honestly, it's not even writing perfect prompts anymore. Your job is just having taste. You look at what got built, and you hold it against the standard that you know of, and you say, not good enough, here's the reference, here's some context, whatever it may be.
13:20You don't manage the engineers, you just manage the manager. It's funny enough, but that's what you knew. Alright.
13:25Now round two, it is going live, and if we just take a look at the screen here, we have four specialist teams all at once. Now this one, this is rebuilding a scoring around the real frameworks that the sales teams are training on. So we have Bance, Med, DPIC, we have NEPQ, and this one is on the persona library, and even a call review tool.
13:44So this one's it's building teams, it's building the leaderboards, and it's also doing the coaching. And this one, I never even asked for. So it found it on the reference product.
13:53We have a hiring flow where a candidate role plays against your AI buyer before you ever interview them, and this one, it frames the whole thesis. So that is four departments, one manager, and zero meetings. All of this, it's not going to stay extremely clean, and honestly, I would say that's good.
14:09You should be seeing a failure, and this worker right here, it just died mid task, so it exited with an error, and if you just watch the manager work from there, it notices this on its own. So it reads the log, it figures out what went wrong, it rewrites the brief, and it relaunches it. So this is self annealing.
14:26Nobody had to wake me up, and that's the difference between running an agent and actually running a real AI team. Alright. Now next up, Fable, it decides all on its own that the product, it needs a test suite.
14:37And it just writes one itself. So we have an automated harness that grades its engineers work completely end to end. Now my manager, it's writing exams for its employees right now.
14:46And then a little bit after that, we just completely ran out of tokens. So I added initially about $50 into this build for just using Soul specifically, but I mean, I was going pretty hard on what I was asking it for, and I was running it on, you know, pretty much max settings.
15:00So to be expected, I had to top up a little bit adding I think I had about another $20 or so. By the way, I'll show you the full bill at the end of this video. Alright.
15:07Now the last thing the manager does before it actually calls this finished is it just tries to break its own product. So it's really just running a security sweep across everything that the second wave had shipped. Now initially, two alarms came back, and for a second they did look pretty scary, where it just dug into both of them and both turned out to be false alarms.
15:27So good. But one of them, it ended up being real. So the transcript parser, it was misreading timestamped exports, and then that was just one last surgical fix, and it ships.
15:36And then from there, we have the receipts. So every AI path, it tested against the live production site. So we have 17 for 17.
15:44Every other path, it was 23 for 23. So I would say this is green across the board. Now ultimately, the total time from the first prompt to a shipped tested product, it was about two hours and twenty minutes, and that again was just about one session.
15:56And when we get back, this is Deal Dojo, which is our new software product, which I really don't intend to be selling or even using much outside of our internal company. But anyways, so this is the headline that it had come up with, and I think it could be a little bit better had I given it some marketing material and some of my methodologies, or just common marketing methodologies.
16:15Anyways, we have the get started here, see how it works, and we have the AI video. Again, this can be improved, but I didn't really provide much of any guidelines or insights to much of anything. So we show or it shows a little bit how it works.
16:29So we have the how it works section with the building your buyer, run the call, get your scorecard from there. A buyer that pushes back like one, turn any real call into the next coaching plan.
16:40So again, just giving insight on everything that it is about. So you can see we have the start free and get black belt through Sensei, talk to sales. Actually really like how that is pitched.
16:52So let's get started. Let's create our account. Now we'll just click create account, and just like that, it automatically logs us in.
16:56I probably would have preferred to have a feature where I had to confirm my email address just to make sure we don't get any bullshit sign ups, but anyways, this is fine. So let's go through the dashboard a little bit. We can see we just have some of the most common features.
17:07So your progress starts here, we can start practicing, we can start your first round, start a role play. We have the new role play inside of right here, call review, personas, and oh, there's actually a few different personas, and the team as well.
17:22So we can set up our teams if you wanted to create one. But let's actually just start a new role play. So now let's just check out the so let's just check out the role plays and see if we could actually get something up and running.
17:31I wanna make sure that the voice is actually working properly. So let's go to, like, Doug Whitman, strategic procurement lead.
17:38And what do you want to sharpen? So let's try to sharpen our objection handling. We'll do proven scenario.
17:46I mean, looks like there's only one saved, and I'm sure we can add some easily. But industry, we can leave that as is, and let's do not too expensive.
17:55We'll I already have a tool. Let's do need to ask my boss as well, And let's start the role play on this.
18:02So first up, you'll notice that it's going to say the text right here. I'll be direct to competitors 50% cheaper on paper. We already have a maintenance analytics tool in place, so I need a compelling reason to keep Optimal in the process, however that's pronounced.
18:14So we could either just go through this right here, or we can go through what's going to be the most beneficial is going to be the voice. Now I'm not exactly sure if it's going to work right now because I'd already said this first text, so let's try this out.
18:27Yeah. So voice recognition lost its connection. Let's try this once again, where it's going to first start out with the voice.
18:32So that is just a small bug, but let's just change a couple things, start the role play. I'll be direct. Your competitor is 15% cheaper on paper.
18:42And unless UptimeIQ can show standardization savings that clearly outweigh switching costs, I can't justify moving this forward.
18:49And then just a second later, we save voice recognition, loss of connection. Text chat still works. So it's right here.
18:54We're just gonna have to go back into the terminal and actually provide all of the changes because obviously it's not good to go right now. So that's one more thing that we have to add to the list to fix before we actually go into production if we were going to be actually deploying this and go to market with it. Moving on, let's see if there's any other features that we want to be exploring.
19:10We can see our session history right here, which is pretty much what we're looking for. Now if you go into our settings, we can actually plug this into Slack. So our team, we actually operate internally using Slack, and I'm sure most other agencies out there do as well as it's one of the best communication channels.
19:26But we could also add anything else in here, like ClickUp, just so it can notify me and create tasks if need be. What else do we have?
19:33Inside of the dashboard, we can create our role plays, which we have just done. We can also build up from scratch. It's actually not gonna look much different, but but for the most part, this is pretty much everything that I'd asked for it to build out, and the only one thing that we need to change is going to be the role play to make sure that it's actually going to get in the conversation, fix that bug, also change the voice, and we could probably just configure Eleven Labs because that voice, obviously, it's extremely robotic, but not that it really matters if it's just used for our sales teams, and it's not gonna be client facing and really try to sell people anything.
20:02And then with the call review, this is where we're going to be able to actually provide any of our specific previous sales calls. So what I would want to do is I would want to go into Fireflies and get all those transcripts and plug it into here. So what I can even do is use Claude code, give it my API from this specific platform where it has built out the API documentation and everything.
20:20We have the curl that we can use to give to Claude code to say, hey, I want you to take all my Fireflies transcripts and everything and put them into this and then score them. Anyway, so that's what the application actually looks like.
20:31But now what does all of this cost? So the engineer sold it, ran on API credits as I had mentioned earlier, and the whole build, both rounds, the test suites, pretty much everything, it was about $80 in tokens. So the manager, Fable, cost me nothing extra because when we ran this build, Fable came included with the Claude Max plan that I have and already paid for on my subscription, and then the engineer invoices by the token.
20:53So the boss is on salary in this specific case if we're trying to take that perspective. And the pricing per table, it explains why the team is shaped this way. So Fable, it is $10 in, $50 out per million tokens.
21:05So this is the most expensive brand on the board. Now SOL, this is $5 in, $30 out, about a half.
21:11So Terra, that's also half of Sol, and Luna, this is a dollar in, six out. So the expensive model things, and the cheaper models, these are typing. And you never want your $10.50 model just hammering out your boilerplate, and you never want your $1 model just making architecture calls that nobody reviews.
21:28So make sure you are delegating responsibly and appropriately. Now if we put this against the receipt, so the real product, this is about $55 per seat per month, And if we have maybe 10 reps, which is kinda what we're getting close to, that's $6,600 a year, every single year.
21:42So mine cost me about $80 just once, and if I had another $10, we can fix everything. We can connect to 11 Labs, get better voices, and just make all the bug fixes, and literally make it perfect and optimized to whatever standard that my team or I want. So this will literally pay for itself before editing this video.
21:59Alright. And before I give you my final verdict on pretty much everything, I just wanted to cover a few different things. Number one, this product has .
22:05So literally every role play call, it is just burning through API credits, same as the product that it competes with. Now, when my balance, it died at the end of the build. Fresh calls, they also died with it.
22:17And that's why every AI moment that you just watched is real footage from the build session. And the reason that I just trust this still works is the test suite. So it came up with 40 test screen on production, and 17 of them are just driving the AI end to end.
22:30Number two is you are paying two companies. So Claude Max for the manager, OpenAI API credit for the engineer, because Soul hadn't rolled out to my ChatGPT login yet, and the API just skips the wait. So if I were you, I would just try to use the subscription for ChatGPT.
22:44Obviously, it's gonna be cheaper. But if you're not worried about cost, which I highly doubt that's the case, then do whatever you want. Now number three.
22:50So the wiring, it's a command line bridge, not just this native feature that you can be using. So I will say that every worker, it starts from a written brief, not the full conversation. So details, they fall through the cracks, and that's not a reason to skip this.
23:03It's literally the reason that the review step even exists. Number four, the benchmarks. They are selective, and that's on both sides of the equation.
23:10So SOLE, it wins the terminal test, and that's about 88.8 on Terminal Bench. In Cloud Fable five, it wins the software engineering one, so that's about 80.3 on SWE Bench Pro, where OpenAI just hasn't published Sol's number whatsoever.
23:24Now when a company skips a benchmark, silencing usually tells you something, so I would say trust the product on your screen, not the leaderboard. But anyways, number five, SOL itself. It overbuilds.
23:36I got features that I never even asked for, and on a paid API, extra work for me, and for you, that's gonna be extra tokens. And beyond that, it misses a lot of things. So for example, the review pass, it caught real defects, and about two of them were very serious, but the manager, it caught all of it before it even shipped.
23:52But nonetheless, it still existed. And none of that changes were actually land on this. It just tells you what you're actually going to be signing up for, and I want you to really know what you're going to be getting from these two models.
24:01But overall, I will say that Soul has been a little bit more rigorous. I've seen other people compare it as like some sort of where it keeps attacking until it finishes the goal, although sometimes it might go a little overkill, whereas Fable it's a little bit more meticulous, and it's going to be a bit more strategic.
24:15Again, that's where it's kind of just being the better manager. Alright. So the verdict, I would say that Sol is genuinely a great engineer, and after a full session of just managing this, I would put it right around Opus level, and it's priced like Opus two.
24:28So it's definitely not Fable. Fable is a tier above, and that gap isn't the code. It's the judgment.
24:34So the planning, the delegating, even the the catching of the security holes, noticing a dead worker, and just re dispatching it, that is the expensive skill, and it's exactly the skill that you want at the top for you or your business or your clients. So don't pick a winner like this or that. I would hire both, Fable manages, sole builds, and you can just hold the standard.
24:55So if you've already got Fable running in Cloud Code, it's about 10 or $20 on an OpenAI API key, and that gets the team just wired up and working today. But again, if you have it on the subscription, highly recommend. As you know, you should just be sticking to that.
25:07So a build the size of this, it was pretty extensive. This one ran me about $80. And if you could only justify one subscription, just run Fable alone.
25:15It can do both jobs, it's just slower, a little pricier per line of code. But as of this week, you can literally hire a manager from one Frontier lab, and an AI engineer from its biggest rival.
25:25So you put them on one team on your own computer, and you can just watch them ship a real commercial product in a single afternoon, if that. Now it's the correct org chart that I would recommend for you or for any complex builds. And if you want to hire the same team, the exact manager, the prompt, the build brief, they're all available for completely free inside of my community.
25:42AI Accelerator, link can be down below in the description. With that being said, thank you guys for watching. I hope you got some value from this video.
25:48If you did, drop a comment, drop a like. I'd really appreciate it. And if there's any other videos that you want me to see uploaded up next, then just drop a comment.
25:55Let me know. Always take a look at every single one of your guys' comments. But thank you guys.
25:59I'll see you in the
The Hook
The bait, then the rug-pull.
Every AI channel filmed the same "Sol vs Fable 5" video this week, so the creator skipped the fight and made them coworkers instead, handing one model the coding and the other the job of managing it.
Named ideas worth stealing.
02:17model
Manager/engineer AI split
- Manager plans, delegates, and reviews; never writes code
- Engineer executes code changes at high volume through a CLI tool
- Manager catches security and logic defects before anything ships
One frontier model is prompted purely as an engineering manager while a second, separately-priced model does 100% of the actual coding, coordinated through written briefs rather than shared conversation context.
Steal forany solo-founder or agency build using two different AI coding tools together
13:38list
Sales scoring frameworks referenced (BANT, MEDDIC, NEPQ)
- BANT
- MEDDIC
- NEPQ
The rebuilt scoring engine was wired to grade AI role-play calls against established sales-qualification methodologies rather than a generic rubric.
Steal forany AI call-scoring or sales-coaching tool
How they asked for the click.
VERBAL ASK
25:50link
“the exact manager prompt and build brief are free inside my community, link in the description”
Soft CTA at the very end after the verdict, pointing to a free community rather than a paid product, a low-pressure creator-monetization pattern; heavier monetization (paid 1:1 agency offer, lead magnet) lives only in the description, never pitched on camera.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
