Blog

Practical insights on AI, RF systems, and technology consulting — no hype, just what works.

Developers who let coding agents write both code and tests say they no longer trust the suite. A Thoughtworks article proposes moving the contract into plain data files that a person can read and a pipeline can protect.

Read more →

A developer's agent wrote to /etc without sudo by using Docker group membership. Google's agent runtimes show where the fix is heading: isolation around the agent, not trust in its manners.

Read more →

Vendors now say a model ID never changes underneath you. The agent built on it still can, through prompts, tools, and the serving stack, so programs need a release check that watches behavior.

Read more →

Amateur recordings of operating spacecraft, with open decoders and a metadata standard, give small teams labeled RF data at no cost. The Tianwen-1 story shows both the payoff and the trap.

Read more →

A campaign on September 1 put shell payloads in issue and pull request titles, and the earlier Trivy compromise shows how far a build pipeline's credentials reach. Flight and ground software teams should treat CI as part of the attack surface.

Read more →

Three satellite autonomy announcements this year look alike in the headlines. Each measured something different, and each left out numbers a program would need.

Read more →

Satellites get the attention, but public audits show the ground systems that command them are where schedules, budgets, and capacity quietly run out.

Read more →

Detecting GPS interference is now cheap and public, but locating the transmitter still comes down to sparse data, a search radius, and someone on the ground with a direction finder.

Read more →

Trending repos tell you what someone built. arXiv tells you what just became possible. Neither tells you what your user is missing. The strongest uncrowded source of product ideas is a century of experiments software does not read.

Read more →

A general model is fluent everywhere and reliable nowhere in particular. What domain knowledge adds to an AI system, what the evidence shows about domain-trained models, and how to tell whether you need it.

Read more →

How to actually assess what a dataset is worth: why quality is a relationship between data and use, the six dimensions to check, and how to value data that feeds a model.

Read more →

What small defense-tech companies need to know about using AI in SBIR proposals this cycle. What's required, what's risky, and how to stay clean.

Read more →

Why Stellabyte exists and what we actually do.

Read more →

A plain-English breakdown of RF digital twins, why they matter for defense and spectrum operations, and what makes them different from generic digital twins.

Read more →