Topics worth your time this issue: a deep dive into engineering development frameworks and how teams build across different industries, an updated market map of physical AI and robotics in 2026, using Claude Code to design test fixtures, Kia's humanoid robot factory plans, and more.
Interesting Lecture Notes: On Hardware Development Frameworks
Hardware development frameworks tend to look different on the surface - consumer electronics runs sequential waterfall scheduling, automotive and highly-regulated industries operate on V-Model framing with requirements traceability baked in, while new-age space startups have shifted toward ‘agile’ iterative development. At the core though, all frameworks aim to solve the same underlying problem: how do you manage integration risk in a physical system where changing something late is expensive?
From the DSpace at MIT, a classic set of lecture notes on product development processes (lecture #11, circa 2003), but more relevant now than ever as teams building complex hardware has expanded well beyond traditional primes and OEMs.
Most engineering teams don’t pick a process so much as inherit one, and what’s interesting is how much the frameworks rhyme across industries despite looking different on the surface. The same variables (build volume, system complexity, degree of vertical integration, regulatory environment) just get weighted differently depending on what’s built.
For example, SpaceX and Anduril treat hardware as learning material; rapid build-test-redesign loops where a failed unit is just the cost of data. Traditional defense primes run highly controlled V-Model flows with frozen requirements and certification-driven verification. Apple and Oura run EVT–DVT–PVT, where architecture is locked early and late changes are prohibitively expensive once tooling begins.
Worth reading as a useful reference regardless of which framework is in play. More recommended reading: the NASA Systems Engineering Handbook.
Interesting Chart: Robotics & Autonomous Systems, 2026
We’ve written a fair amount about robotics over the past few issues (hard not to, given how much is moving in terms of engineering development!). This is our annual map update, now rebranded as “Physical AI, Robotics, & Autonomous Systems” to reflect where the field sits in 2026.
The main addition is the intelligence layer, which didn’t really exist as a discrete category a couple years ago.
More From Around the Web
Things worth reading that didn’t fit anywhere else this week:
- Using Claude Code as a test fixture/jig-making tool. Feed it sketches, and get a file ready to 3D print via FreeCAD.
- A neat real-world demo of PID control on a differential drive robot. Also: turns out Instagram Reels might be one of the better formats for visualizing fundamental control theory.
- A full die casting cycle, from start to finish.
- The cyberdeck aesthetic: a whole subcommunity building gorgeous, chunky shells around old laptops or a Raspberry Pi.
Manufacturing & Startup News
More leftovers from our weekly research:
- Kia plans to deploy Boston Dynamics’ humanoid robots in its US factories by 2029 and release its first ‘software-defined’ vehicle (i.e. features can be remotely updated like a smartphone) by 2027, as part of a broader technology and electrification push.
- Askari Defense raised a $1.7M pre-seed round to build low-cost drone-intercepting systems from investors including Knoll Ventures, Overline, and Georgia Tech Foundation.
- Sonibel raised $1.6M in pre-seed funding to develop an acoustic sensor that mounts to a welding torch and uses ML to catch defects (including those inaudible to the human ear) for better weld quality control.
- Terawatt Infrastructure is raising up to $1.5B in equity and debt to build a national network of charging depots for autonomous EV fleets like Waymo, pivoting from its earlier focus on electric truck charging.
- Merino Energy launched an all-in-one heat pump that installs in one hour for $3,800 total, with a SEER2 rating of 15.2; above the federal minimum of 14.3 but below standard mini-splits (18-25; higher = lower electricity bill). The core bet is that simpler design enables flat-rate pricing and removes the costly quote process that was itself a bottleneck to heat pump adoption.