Channel notice
“
Make AI your superpower
Make AI your superpower
Spend nothing. Get AI
running.
Every pitfall I hit, every number I measured, every tool I built — all handed to you.
Local deploymentFree resourcesSelf-made softwareReal data, failures included
Channels
Find me here
Different platforms, same HyphenTech.
Home turfBilibiliFull tests
Long videos premiere here: install, speed, cost and every pitfall, start to finish.
Watch full videos →YouTubeGlobal mirrorFull videos mirrored for viewers outside China.
Watch on YouTube →WeChat ChannelsVertical quick takesKey steps squeezed into vertical clips — short enough for an elevator ride.
Search "黑粉科技" on WeChat →WeChat Official AccountLong-form write-upsFull tutorials, settings tables and download links — worth saving for when you actually build it.
Search "黑粉科技" on WeChat →Three tracks
Start here
Run it locally, use it free, build it yourself — one pick from each so you don't get lost in the archive.
Local deploymentLocalBrain Network Security Test: Locked…
LocalBrain Network Security Test: Locked in an isolated sandbox, it doesn't run by default, only released after authorization
LocalBrain provides workspace security review and authorized network checks. This article records the real reproduction, repair, and local test site o…
Free resourcesLTX-2.5, free local run, with effects co…
LTX-2.5, free local run, with effects comparable to blockbuster films
LocalBrain 1.4.8 connects to the second local video engine, LTX-2.5: Write a quoted li…
Self-made softwareHyphenScreen: The best screen recording …
HyphenScreen: The best screen recording software in the Eastern Hemisphere, integrating screen recording and editing, and automated AI editing
After recording, you cut it directly, leaving it to AI in one sentence: 28 minutes of …
Self-made software
My tools
When I hit a problem I look for a tool; when there isn't one, I write it. Partly out of stubbornness.
Built by me · stable releases
LocalBrain
Local AI toolbox
Latest 1.5.7Release notes ↗
I gathered the local models, transcription, image and video generation, documents and MCP scattered across my Mac into one local workbench that downloads, starts, chats…
Runs locallymacOSActively updated
Built by me · stable releases
HyphenBox
Free API radar
Latest 1.0.0Release notes ↗
HyphenBox has officially launched, starting from version 1.0.0, and offers installation packages for Mac, Windows, and Linux. Early screenshots and detection numbers in…
ReleasedmacOS / Windows / LinuxFree download
Built by me · stable releases
ScreenLex
Learn English from your own films
Latest 1.0.5Release notes ↗
Local film and TV English learning tool: scan local SRT/ASS subtitles, organize offline into an advanced English vocabulary database graded by exam, accompanied by sent…
Works offlinemacOS / WindowsActively updated
Built by me · stable releases
HyphenScreen
Screen recording + smart editing
Latest 1.0.8Release notes ↗
The worst thing a screen recording can do is show what it should not. HyphenScreen 1.0.2 finds sensitive text on your own computer, keeps the mosaic on it through zooms…
FreeRuns locallymacOS / Windows / Linux
Archive
More hands-on tests
Looking for a specific fix? Search — the search box remembers better than I do.
Local deploymentMac, Windows, Linux—which is the main de…
Mac, Windows, Linux—which is the main device in the AI era?
For AI computers, Mac should buy shared memory and low maintenance, Windows should be compatible with graphics card options, Linux should be…
Read more
Free resourcesOn September 23, 2026, how many free AI …
On September 23, 2026, how many free AI APIs are still available?
17 old entry points were reviewed one by one, and 4 candidates were reviewed; 3 callable platforms were screened based on current non-empty …
Read more
Local deployment × Self-made softwareIf a local model can't handle long tasks…
If a local model can't handle long tasks, it's almost certainly not the model's fault
All 12 sets are delivered, but repeating the same problem is 15 times worse: long tasks get stuck in the loop outside, not in the model.
Read more
Local deployment × Self-made softwareQwen-Image-2.1: Lags three times before …
Qwen-Image-2.1: Lags three times before the image appears
31 GiB weight, 20 steps and 69 seconds on the M5 Pro; Chinese calligraphy correctly writes four characters. Connected to LocalBrain and hit …
Read more
Local deploymentThe Mac native engine, which claims to b…
The Mac native engine, which claims to be twice as fast, was only 44% faster in Chinese
Splash generated Qwen 3.8-27B code on my M5 Pro at 79.3 tok/s, 5.11 times faster than the naive MLX with the same weight; Switching to Chine…
Read more
Free resourcesThis model doesn't write a single word, …
This model doesn't write a single word, only answers "Which one to choose?"
Jev doesn't generate text, only makes probabilistic judgments based on status. The official multiplier is impressive, but they put the exagg…
Read more
Local deployment × Self-made softwareTernary Bonsai 2 27B on Mac: 6GB running…
Ternary Bonsai 2 27B on Mac: 6GB running 27B Reasoning, paths, commands, and pitfalls
PrismML reduced 27B reasoning to 5.93 GB, M5 Pro ran 27.7 tok/s, still retaining 98.2% of Qwen 3.8's raw capability. Clearly wrote down the …
Read more
Free resources × Self-made softwareHyphenScreen: a free screen recorder tha…
HyphenScreen: a free screen recorder that edits, redacts and checks its own output
The worst thing a screen recording can do is show what it should not. HyphenScreen 1.0.2 finds sensitive text on your own computer, keeps th…
Read more
Local deployment × Self-made softwareCan the Local 27B Model Operate Blender?…
Can the Local 27B Model Operate Blender? I brought Bilibili's method into LocalBrain and ran it five times
Using local Qwen3.8-27B modeling with blender-mcp, then testing it in LocalBrain for five rounds: the model can be done step by step, but sl…
Read more
Local deployment × Free resourcesOld notebooks that gather dust can start…
Old notebooks that gather dust can start a business
Local AI is not a developer's toy. A comprehensive four-piece set of models, warehouses, software, and workflows is explained, plus three id…
Read more
Local deployment × Self-made softwareWhich local models can 8GB run? I tested…
Which local models can 8GB run? I tested 11 models, and 6 can fit in 8K context
Bringing 11 low-memory local models released this year to the local machine, using LocalBrain's built-in llama.cpp unified memory, speed, an…
Read more
Local deployment × Self-made softwareLocal H3 multi-reference image test: all…
Local H3 multi-reference image test: all 4 images used, 3-second video runs for 10 minutes
LocalBrain 1.2.65 can run MiniMax H3's multi-reference plates (REF2VA) locally, with up to 9 images at once. On a 64 GB M5 Pro, I used 0, 1,…
Read more
Local deployment × Self-made softwareI had the local model run for two consec…
I had the local model run for two consecutive days to write five web games: 1 delivered, 3 ran, and 1 abandoned
Qwen3.8-Flash-Next on a 64GB Mac ran five offline web challenges: 1 delivered, 3 executable, 1 abandoned. Every timeout, crash, and long con…
Read more
Local deployment × Self-made softwarellama.cpp 0.4:64GB Is the hard wall loos…
llama.cpp 0.4:64GB Is the hard wall loose?
llama.cpp 0.4 rewrote the memory boundaries of large models with lazy loading, streaming quantization, and Apple RDMA. But LocalBrain's 64GB…
Read more
Local deploymentCan a 1.4GB small model really act as a …
Can a 1.4GB small model really act as a local agent?
MiniCPM5-2B packs 130,000 contexts, agent training, and official GGUF/MLX into 2.52B parameters; The real highlight isn't the benchmark scor…
Read more
Local deploymentIs local AI safe just by opening localho…
Is local AI safe just by opening localhost?
Just listening to 127.0.0.1 doesn't mean browsers can't encounter it. Three recent official security announcements plus a set of completely …
Read more
Local deployment × Free resourcesNex-N2.5 is free, can a 64GB Mac run it?
Nex-N2.5 is free, can a 64GB Mac run it?
35B multimodal agent is open source, and online endpoints are currently free; Whether Mac capacity is sufficient means runtime is already co…
Read more
Local deployment × Free resourcesOmarchy: Free Linux for AI Mastery
Omarchy: Free Linux for AI Mastery
From installation and shortcuts to programming assistants and local models, it explains Omaly's features and getting started with it.
Read more
Local AI · measured data
AI lab
Hardware, local models, agents and a 48-hour release radar — click, look things up, use them. (Pages are in Chinese.)
◫AI hardware guideLocal AI devices compared⌘Local LLM guideOllama · LM Studio · GGUF△MLX model testsIn-depth local model benchmarks on M5 Pro◉Release radarMajor releases within 48 hours≠AI agent showdownThree agents, head to head
Bonus drops
Side projects
Not the main act, but they're built — take whatever helps.