Creative Briefs

Relume Base44 Genspark 3.0
Resolve before any of these ship.
Genspark 3.0 is tracked as "Dedicated 1/1" but every email prices it as an integration — "The price for an integrated video is $15,000" / "intergation 15kusd about Genspark( 90s and transfer fee is already included)". Rate card: integrated $15k, dedicated $30k. A 90-second segment at $15k is an integration. $15,000 gap.
Base44 per-video rate is unconfirmed — $27k on the roster vs $90k-at-list in the meeting brief. Do not invoice until Tal Karshi confirms.
Relume — $5,000 of $20,000 invoiced (INV-10, deposit on video 1 only).

Relume — Integrated, video 1 of 2

Hard constraint, already conceded by Jack and locked: titles must NOT name Relume, and no Relume logo on the thumbnail — he refused both on performance grounds and the terms held. Naming Replit is fine. Deliverables per Dan Anisse, 2026-08-04: "First minute mention + platform shown in every video / Pinned link on each." Zhara manages script approval for both videos.

The concept — "The First Ten Minutes"

Premise: The part of a client website nobody films is the part that decides whether it works — the structure, before a single pixel or line of code. Jack scopes a real client site in the first ten minutes, then lets Replit build what that structure produced.

Why this one: it satisfies the first-minute mention honestly, because the tool genuinely belongs at minute one. It never needs the brand named in the title, which is the locked term. And it leaves the obvious sequel intact for video 2.

0:00–0:20
Hook. A finished client site, then a hard cut to a blank page. "This is the ten minutes that decided everything."
0:20–1:00
The real brief. An actual client request, read out as received — vague, contradictory, no structure. Platform on screen inside the first minute.
1:00–3:30
Structure first. The sitemap builds. Jack argues with two of its choices and keeps one. This is the segment that carries the mention.
3:30–6:00
Wireframes. Page by page, still grey, still no design. The point being made: the hard thinking is already done.
6:00–12:00
Replit builds it. Structure in, working site out. Jack does no design work on camera — that is the proof.
12:00–end
The price. What an agency charges to scope and ship this, against the clock. Pinned link called out.

Video 2 (for later, don't spend it here): the same workflow against a deliberately harder brief — multi-page, real content, a client who changes their mind — so the sequel earns its own hook instead of repeating this one.

Titles

★ = recommended. Every option below is unbranded, per the locked terms.

TitleIn-image text
★ The First 10 Minutes of Every $10,000 Website I BuildFIRST 10 MIN
I Built a $10,000 Client Website in Replit (Zero Design Skills)$10,000
How I Scope $10,000 Websites in 10 Minutes (Before Any Code)10 MINUTES
Replit + One Secret Weapon = $10,000 Client WebsitesSECRET WEAPON
Zero Design Skills to $10,000 Client Website (Full Workflow)$0 SKILLS
Relume concept thumbnail
Existing render — fully unbranded per the locked terms. Grey wireframe sitemap against a finished site.

Base44 — Concept outline

The only stated brand rule, from Tal Karshi, 9 Jul 2026: "The only adjustment I'd like to make is to ensure Base44 is more front and center in the video. Since the goal is to showcase base44, we need to ensure the platform's role in the process is prominent throughout the narrative." No title or logo restriction exists. The brand declined to supply a concept: "we don't have any specific ones in mind... We prefer to leave that up to your team."

The concept — "A Real Client Brief and a Clock"

Premise: A genuine client brief, a visible timer, and Base44 doing the build end to end — with the model picker treated as a decision Jack makes on camera rather than a feature he lists.

The authenticity problem, handled head-on: Jack has said on camera that he defaults to Claude Code and named Base44 as a tool he does not use. Pretending otherwise reads as bought. Saying it out loud in the cold open and then being changed by the result is the strongest version of this video — and it keeps the platform prominent throughout, which is exactly what Tal asked for.

0:00–0:45
Cold open. "I've said before I default to Claude Code. So I gave this a real client brief and a clock." Timer starts.
0:45–2:00
The brief. A real request with real constraints — scope, users, what it has to do on day one.
2:00–5:00
The model decision. The picker comes up. Jack reasons out loud about which model fits this job and commits to one. Nobody else's video has this beat.
5:00–18:00
The build. Front end, data, auth, deploy. Jack reacts to what it does unprompted and to what he has to correct.
18:00–22:00
Stress it. Change the brief mid-build the way a client would. This is where a builder either holds up or doesn't.
22:00–end
The verdict. Timer, what an agency bills for the same scope, and an honest read on where it beat his own stack and where it didn't.

Screen accuracy note: the first thumbnail attempt invented outdated model names on screen (Sonnet 3.7, GPT-4o, Gemini 2.5, Llama 4). The sponsor would have caught it. Whatever model names appear on camera must be the real current roster — confirm with Tal before shooting.

Titles

★ = recommended. No naming restriction — Base44 can and should appear in the title.

TitleIn-image text
★ Base44 Built a $9,000 App in 24 Minutes (I Picked the Model)$9,000
I Tested Base44 Against My Own Stackvs CLAUDE CODE
Base44 Lets You Pick the Model — Here's Which One WinsWHICH MODEL?
I Stopped Paying a Developer $9,000 (Base44)$9,000 SAVED
Base44 + Claude = A Real Client App in One SittingONE SITTING
Base44 concept thumbnail
Existing render — brand orange verified, model-selector buttons deliberately blank so no outdated model name appears.

Genspark 3.0

Format unresolved — settle this first. Tracked as "Dedicated 1/1", but priced and described throughout as a 90-second integration at $15,000. Dedicated list is $30,000. The brief below is written as the integration the emails describe. If it is genuinely dedicated, the fee is wrong and the runtime below is far too short.
Brand rules, verbatim. Title: "with it being an integration, Genspark can't be in the title." Model firewall, stated twice: "The Fable 5 model cannot be used in Genspark Design... when it comes to the Genspark portion, please downplay 'which model' and instead focus on 'what deliverables can be produced.'" Required framing: "Hermes handles conversation and orchestration, but to produce actual deliverables (decks/websites/videos), you use Genspark Design. The differentiator is: others give you answers, Genspark gives you finished products." Approval before publishing is mandatory; no revisions after publish.

The concept — "Answers vs. Shipped"

Premise: The agent gives a brilliant answer. The answer is not the work. Hard pivot into Genspark Design producing a finished brand kit, a promo video and a live landing page — ending on the publish click.

This is the brand's own differentiator turned into structure rather than a claim. The pivot line does the whole job: "That's a great answer. But an answer isn't a deliverable." Everything before it is Fable-5-safe; everything after it is a Fable-5-free zone, which satisfies the firewall by construction instead of by remembering.

0:00–0:20
Segment 1 — the answer. Hermes produces a genuinely good plan for a client brand. Models may be discussed freely here.
0:20–0:30
The pivot. To camera: "That's a great answer. But an answer isn't a deliverable." Model talk stops here and does not resume.
0:30–1:00
Brand kit. Logo, palette and fonts from a single prompt. No mention of which model — only what came out.
1:00–1:20
Promo video. Same style, produced directly. The brand specifically asked for this over a deck demo.
1:20–1:30
Landing page and publish. Generate, Present to preview, then Build It. The publish click is the payoff shot — hold it.

Conflict to fix before sending: Jack's existing drafted brief ("Hermes + Fable 5 = Unstoppable", Case 2 = "Fable 5 builds a deck from a prompt") directly violates the firewall and would fail approval as written. Unknown, do not invent: nothing on file defines what "3.0" is as a version — get it from Eezycollab rather than writing around it.

Titles

★ = recommended. Zero options name Genspark, per the brand rule. "Hermes" in a title needs Eezycollab sign-off.

TitleIn-image text
★ Most AI Gives You Answers. This One Ships.ANSWERS vs SHIPPED
One Client Call Became a Deck, a Site, and an Invoice1 CALL, 3 ASSETS
I Built a $10,000 Brand Kit With One Prompt$10K
A Full Brand Identity in 8 Minutes8 MINUTES
The Only Thing I Did Was Upload the TranscriptONE UPLOAD
Genspark concept thumbnail
Existing render — brand blue verified from genspark.ai/brand. Whether the wordmark may appear inside the image was never addressed by the brand; the title rule is title-only.