How we work, in our own words.
Long-form writing on hardware, AI, and how the ByrdByte team builds. One post a month. No clickbait, no AI-generated filler. Either we have something useful to say or we don't post.
Hardware, photography, real-world setups, things we tested.
Your MacBook is out of storage. Here's the honest fix.
"Your disk is almost full" is the most-Googled MacBook problem there is. This is every real way to reclaim space — what actually works, what only buys you a week, and the one change that ends the treadmill for good.
Case X for editors: 4K multicam without the proxy round-trip
A ten-minute TikTok ingest can hit 80 GB. A wedding edit lands at 800 GB. Editors who shoot 4K ProRes are running out of disk faster than they're running out of footage. Here's the math, and the drive.
Why Case X is the storage your beats deserve
Sample libraries are now 2 TB out of the box. Project files are 30 GB before mixdown. Your MacBook is full again. A field note for producers tired of paying the Apple storage tax.
The cloud is running out of power, and AI keeps asking for more
Hyperscale data centers now burn megawatts of electricity per facility — and grid hookups have 24-month waitlists. Why AI broke the cloud's energy budget, and what comes after.
We tested five mini PCs for running local AI. Here's what we'd actually buy.
A real walkthrough of which mini PCs handle on-prem AI well — including the M4 Mac Mini, Beelink SER9 Pro, MSI CUBI NUC AI, ACEMAGIC S3A — which one we'd put in a small law firm or design studio, and where the Mac Mini wins vs loses. No affiliate links, no hype.
How we use AI inside the company, and how ByrdDrive thinks.
Model weights, datasets, and the engineer who can't put them on AWS
A 70B model is 140 GB. A fine-tune dataset is 200 GB. The contract says privileged data never leaves the office. The engineer's MacBook is 1 TB. There's a drive that fixes this.
Auto-extracting invoices into spreadsheets, with a model that never leaves the building
How ByrdDrive turns a folder of scanned PDFs into a structured table — vendor, amount, due date, category — using a local AI model. The actual pipeline, not the marketing version.