Hype Check · Ras Mic

Worth your time?

Name: Is "AI is expensive, which subscription should you get?" worth watching? Hype 34/100 | Clip'd
Item: AI is expensive, which subscription should you get?
Rating: 66
Author: Clip'd

AI is expensive, which subscription should you get? by Ras Mic

Watch on YouTube

Hype score · lower is better

34/100

Useful but promotional

The video mostly teaches practical, budget-focused subscription choices and actionable settings, with a light sponsor and a couple self-promos.

Substance65

Proof45

Pitch pressure28

Claim inflation30

The 15-second version

Post it, send it, settle the argument.

Download clip

Receipts

Why this score

Scored from the transcript, summary, and output signals.

Title deliveredtitle asks which subscription to pick; the video compares Codex, Claude, Cursor, and Devon (4) and clearly recommends Codex with pricing and usage tips

Skip the video

The short version

Kept by Clip'd

Ross compares Codex, Claude (Cloud Code), Cursor, and Devon for people on a budget, weighing model quality, subsidies, harness features, and pricing. He concludes Codex offers the best value due to heavy subsidies, strong GPT models, and useful harness features (plugins, in-app browser, computer use), and gives concrete model-and-setting recommendations plus a tactic to get your employer to cover a subscription.

Keep the lesson

What it teaches

Cost-quality tradeoff as a decision model: choose the best model that delivers 'enough' accuracy for the task, not the absolute best. Apply this when token costs scale with usage (e.g., large codebases, heavy testing). The common mistake is defaulting to the highest-tier model (Ultra/Fable Max) for routine work and bur
Subsidy leverage principle: prefer providers or harnesses that frequently subsidize compute (resets, free quotas) when you expect sustained heavy usage. Use this for exploratory experimentation or CI-like automated workflows. The common mistake is ignoring how subsidy cadence affects effective cost per token and overpa
Harness-feature fit: pick a harness by workflow features you actually use (in-app browser, computer use, thread management, plugins) rather than raw model labels. Apply this when your work depends on automation, local/cloud switching, or integration with services. The common mistake is selecting a harness solely on mod

Jump straight there

Key moments

2:21Fable vs GPT-5.6 verdict: Ross names Fable 5 the stronger model but says GPT-5.6 Soul is the better cost/utility choice for most tasks.
4:21Subsidy advantage: Ross demonstrates how OpenAI/Codex provides frequent 'resets' and subsidized compute, giving Codex a cost edge.
7:44Harness winner nuance: Ross calls Cursor the best harness UX but still crowns Codex the winner overall because of model+subsidy balance.
10:30Pricing posture: Ross runs through concrete subscription tiers (Codex Pro 5x $100/mo, 20x $200/mo) and frames $200/mo as high value if you cancel other subs.
14:54Slashgoal demo: he shows the in-app browser + agent workflow (slashgoal) to optimize a site to Lighthouse 100, illustrating automated, long-running agent work.
18:20Token management warning: Ross emphatically warns "Don't use Ultra" because it guzzles tokens and gives poor cost/value.

Lines worth keeping

Quotes

AI is incredible, incredibly expensive0:00
Fable 5 is the better model.2:21
Codex is the one.10:13

Clip'd

Skip the video. Keep the lesson.

Score your next video free

The Hype Board