Creative Automation / Foundation

Colibri Tested Locally! Run DeepSeek V4 (284B), GLM-5.3, Kimi K3 & More on 24GB RAM!

This video demonstrates how Colibri runs a 284-billion-parameter DeepSeek V4 Flash model locally on a 24 GB MacBook Air by keeping shared layers in RAM and streaming mixture-of-experts weights from SSD. It also covers the four-command setup and measures the practical cost of that approach in latency, cache misses, and disk reads.

AI WITH Rithesh4 minTranscript found

Quick learning frame

Read this before watching.

Creative automation accelerates production while keeping human taste in brief, source selection, generation, editing, and critique.

New playlist item from AI WITH Rithesh; queued for transcript-backed review, topic mapping, and a practical learning artifact.

Skill you build: The ability to assess whether an oversized mixture-of-experts model can run on limited local hardware and identify the RAM, cache, SSD, and speed tradeoffs involved.

Watch for the shift from claim to mechanism. The learning value is the point where the transcript reveals a repeatable action, tool boundary, context move, review habit, or artifact.

Concept diagram

Where this video fits.

01Brief
02Source material
03Generation
04Selection
05Edit
06Taste review
07Reusable recipe

Deep lesson

Turn this video into working knowledge.

602 cleaned transcript words reviewed across 184 timed caption segments.

Thesis

Colibri Tested Locally! Run DeepSeek V4 (284B), GLM-5.3, Kimi K3 & More on 24GB RAM! teaches a practical creative automation move: This video demonstrates how Colibri runs a 284-billion-parameter DeepSeek V4 Flash model locally on a 24 GB MacBook Air by keeping shared layers in RAM and streaming mixture-of-experts weights from SSD. It also covers the four-command setup and measures the practical cost of that approach in latency, cache misses, and disk reads.

The goal is not to remember the video. The goal is to extract the operating principle, tie it to timestamped evidence, test how far the claim transfers, and make something reusable.

0:21

Experts From SSD

“out last week. The reason this works is the mixture of experts design. The model is huge, but each token only passes through a handful of its experts. Colibri keeps the shared layers in RAM and leaves the...”

Colibri makes the 284-billion-parameter model fit by holding shared layers in RAM while leaving experts on SSD; a router loads only the experts needed for each token and caches frequently used ones. This mixture-of-experts design avoids loading the full model into memory, but shifts the bottleneck to disk reads. Draw a three-part flow from router to SSD-resident experts to RAM cache, and annotate what happens for a cache hit versus a cache miss.

0:55

Four-Command Setup

“commands. Install Xcode's command-line tools, install libomp, git, and Python with Homebrew, clone the repo, and run start-here.sh. That script looks at your RAM, disk, and CPU, recommends a model, builds the engine, and downloads the weights. If...”

On macOS, setup consists of installing Xcode command-line tools and Homebrew dependencies, cloning Colibri, and running start-here.sh. The script inspects RAM, disk, and CPU, recommends a model, builds the engine, downloads resumable weights, and accepts --model when the user wants a specific model. Write a setup checklist that includes the required macOS tools, repository clone, start-here.sh, the optional --model flag, and the resume behavior for interrupted downloads.

2:33

Disk Sets Speed

“For that run, it read 268 GB from the SSD. That's more than the size of the whole model. Only about 37% of expert requests were served from the cache. The rest came off the disk, and that's...”

The DeepSeek test answered the sheep puzzle correctly but produced only about 0.36 tokens per second: 118 tokens took roughly 5.5 minutes, while the run read 268 GB from SSD and served only about 37% of expert requests from cache. More RAM for caching or a faster drive would improve performance, but this configuration is a feasibility demo rather than a practical daily-work setup. Record the test's model size, RAM peak, SSD reads, cache-hit rate, and tokens per second, then identify which measurement most directly explains the slow generation.

01

Brief

Start with this video's job: This video demonstrates how Colibri runs a 284-billion-parameter DeepSeek V4 Flash model locally on a 24 GB MacBook Air by keeping shared layers in RAM and streaming mixture-of-experts weights from SSD. It also covers the four-command setup and measures the practical cost of that approach in latency, cache misses, and disk reads. Treat "Brief" as the outcome you are trying to make visible, not a topic label. Anchor it to 0:21, where the video says: “out last week. The reason this works is the mixture of experts design. The model is huge, but each token only passes through a handful of its experts. Colibri keeps the shared layers in RAM and leaves the...”

02

Source material

Use "Source material" to locate the part of the creative automation mechanism the video is demonstrating. Ask what changes in your real setup if this claim is true. Anchor it to 0:55, where the video says: “commands. Install Xcode's command-line tools, install libomp, git, and Python with Homebrew, clone the repo, and run start-here.sh. That script looks at your RAM, disk, and CPU, recommends a model, builds the engine, and downloads the weights. If...”

03

Generation

Turn "Generation" into the reusable artifact for this lesson: A creative production board with source inputs, prompt recipe, selection criteria, edit pass, and taste-review checkpoints. This is where watching becomes something you can inspect and reuse.

04

Selection

Use "Selection" as the application surface. Decide whether the idea touches a browser flow, a local file, a model choice, a source document, a UI, or a review step.

05

Edit

Use "Edit" to prove the lesson. The evidence should connect back to the video title, transcript anchors, and a concrete output, not a generic best-practice claim.

06

Taste review

Use "Taste review" to carry the idea forward: save the prompt, checklist, diagram, or operating rule that would make the next agent run better.

07

Reusable recipe

Connect "Reusable recipe" to Colibri Tested Locally! Run DeepSeek V4 (284B), GLM-5.3, Kimi K3 & More on 24GB RAM! by naming the claim, the evidence, and the artifact it should produce.

Example

Source-backed artifact packet

Convert the video into a scoped artifact request that includes the transcript claim, mechanism, acceptance criteria, and proof. The output should be a creative production board with source inputs, prompt recipe, selection criteria, edit pass, and taste-review checkpoints..

Example

Creative automation proof brief

Separate what the speaker claims, what the demo actually proves, and what still needs outside verification before you adopt the creative automation pattern.

Example

Teach-back module

Transform the lesson into a definition, a Brief -> Source material -> Generation -> Selection -> Edit -> Taste review -> Reusable recipe diagram, one misconception, one practice exercise, and a check-for-understanding question.

Do not learn it wrong
  • Treating the title as the lesson without checking what the transcript actually says.
  • mistaking novelty for quality
  • no source/brief discipline
  • shipping generated media without taste review
  • Letting the lesson drift into generic content advice.
  • Letting the lesson drift into tool hype.
  • Letting the lesson drift into creative output without selection criteria.

Transcript-derived moments

Use timestamps to study the actual video.

Quality check

Do not count this as learned until these are true.

01

State the transcript-backed claim in your own words: This video demonstrates how Colibri runs a 284-billion-parameter DeepSeek V4 Flash model locally on a 24 GB MacBook Air by keeping shared layers in RAM and streaming mixture-of-experts weights from SSD. It also covers the four-command setup and measures the practical cost of that approach in latency, cache misses, and disk reads.

02

Explain the practical stakes without hype: New playlist item from AI WITH Rithesh; queued for transcript-backed review, topic mapping, and a practical learning artifact.

03

Map the idea onto the Brief -> Source material -> Generation -> Selection -> Edit -> Taste review -> Reusable recipe sequence and name the weakest link.

04

Produce the artifact and include the evidence that proves it: A creative production board with source inputs, prompt recipe, selection criteria, edit pass, and taste-review checkpoints.

Put it into practice

Give this grounded prompt to Codex or Claude after watching.

You are helping me turn one specific YouTube video into real, durable learning.

Source video:
- Title: Colibri Tested Locally! Run DeepSeek V4 (284B), GLM-5.3, Kimi K3 & More on 24GB RAM!
- URL: https://www.youtube.com/watch?v=TRWgBxVAb1M
- Topic: Creative Automation
- My current learning frame: Given a laptop's RAM and storage limits, sketch a Colibri deployment plan for an oversized mixture-of-experts model and predict whether cache capacity or SSD throughput will dominate its generation speed.
- Why this matters: New playlist item from AI WITH Rithesh; queued for transcript-backed review, topic mapping, and a practical learning artifact.

Transcript anchors from this exact video:
- 0:21 / Evidence 1: "out last week. The reason this works is the mixture of experts design. The model is huge, but each token only passes through a handful of its experts. Colibri keeps the shared layers in RAM and leaves the..."
- 0:55 / Evidence 2: "commands. Install Xcode's command-line tools, install libomp, git, and Python with Homebrew, clone the repo, and run start-here.sh. That script looks at your RAM, disk, and CPU, recommends a model, builds the engine, and downloads the weights. If..."
- 2:33 / Evidence 3: "For that run, it read 268 GB from the SSD. That's more than the size of the whole model. Only about 37% of expert requests were served from the cache. The rest came off the disk, and that's..."

Video-aware target:
- Prompt lane: Creative automation
- Mechanism to extract: Extract the creative production loop, especially where the human keeps taste, selection, and final judgment.
- Artifact to produce: A creative production board with source inputs, prompt recipe, selection criteria, edit pass, and taste-review checkpoints.
- Artifact must include: brief; source inputs; generation recipe; selection criteria; edit/review checkpoint

Your task:
1. Use the transcript anchors above as the primary source packet. If you add outside context, label it clearly as outside context and keep it secondary.
2. Create a source-check table with columns: timestamp, claim, transcript support, what the demo proves, confidence, and what still needs verification.
3. Extract the actual teachable mechanism from the video: Extract the creative production loop, especially where the human keeps taste, selection, and final judgment. Do not invent claims that are not supported by the title, lesson frame, or transcript anchors.
4. Build a reusable learning artifact: A creative production board with source inputs, prompt recipe, selection criteria, edit pass, and taste-review checkpoints.
5. Include:
   - a plain-English definition of the core idea
   - a diagram or structured model using this sequence: Brief -> Source material -> Generation -> Selection -> Edit -> Taste review -> Reusable recipe
   - answers to these source questions: What asset is being produced? | What inputs and tools drive it? | Where does human taste intervene?
   - 3 concrete examples that apply the video idea to real agentic work, such as Claude-generated video campaign; image-to-site workflow; voice or video editing loop
   - 2 failure modes the video helps prevent, chosen from the transcript evidence and these likely risks: mistaking novelty for quality; no source/brief discipline; shipping generated media without taste review
   - a checklist for the next real workflow, focused on: brief, inputs, generation, selection, critique
   - one practical exercise with a clear done signal: Build one reusable creative recipe and define what would make the result rejectable.
6. Add a "learning transfer" section: what changes in my workflow tomorrow if I actually learned this?
7. Add a "source check" section that cites which transcript anchor supports each major takeaway.

Quality bar:
- Make this specific to "Colibri Tested Locally! Run DeepSeek V4 (284B), GLM-5.3, Kimi K3 & More on 24GB RAM!", not a generic Creative Automation essay.
- Anchor each creative step to transcript evidence about inputs, model/tool choices, iteration, editing, or critique.
- Prefer operational examples, failure modes, and reusable artifacts over broad definitions.
- Call out uncertainty instead of smoothing over weak evidence.
- Avoid these generic drifts: generic content advice; tool hype; creative output without selection criteria.
- If evidence is weak or missing, stop and say what transcript segment or timestamp needs review instead of guessing.
- Finish with a concise artifact I could paste into my learning app.

Misconceptions

What to stop believing.

Creative AI removes the need for taste.

It increases the need for taste because output volume explodes.

The best prompt is enough.

References, critique, iteration, and post-production matter just as much.

Practice studio

Learning only counts when you make something.

01

Transcript evidence map

Separate what the video actually says from what you already believe about the topic.

3 source-backed takeaways with timestamps, confidence, and a transfer note.
02

One useful artifact

Apply the video to a real workflow and produce a creative production board with source inputs, prompt recipe, selection criteria, edit pass, and taste-review checkpoints..

A reusable artifact with a done signal and one verification step.
03

Creative automation teach-back card

Explain the creative automation mechanism to someone who has not watched the video yet.

A 90-second explanation, one diagram, one example, and one misconception to avoid.

Recall check

Answer first, then reveal — without rewatching.

How does Colibri run a model far larger than the MacBook's RAM?

What does start-here.sh do after the prerequisites and repository are installed?

What evidence showed that SSD access was the main speed limit in the DeepSeek test?

Source shelf

Use the video as a doorway, then verify with primary sources.

ReadingComfyUIwww.comfy.org/ReadingAffinityaffinity.serif.com/