Topic
Everything filed under Engineering, newest first.
RSS · JSON · All topics
Test/verification technique effectiveness with Astra
Dan Luu re-runs his agentic testing eval with GPT-6 Astra, finding the model capable enough to need a harder benchmark than implementing Zstd from its RFC, and that agents now apply test and verification techniques such as TDD, property-based testing and formal methods more effectively than with GPT-5.6 Sol.
19 min · 4,392 words
How the Claude Code team thinks about verification
Addy Osmani explains how the Claude Code team, where ~90% of code is written and reviewed by Claude, uses end-to-end, property and differential tests, a repo-local verify skill and independent verifier agents to let coding agents prove their work, and how it keeps those verifiers honest.
10 min · 2,352 words
Artist Niklas Roy documents building a working mechanical pendulum clock almost entirely from empty drink cartons, BBQ skewers, paper straws, string, and hot glue, covering the escapement, gear train, design tools, and lessons learned in 'trash bin horology'.
10 min · 2,320 words
Why get an expensive Pi when I have an old phone?
A maker compares building a camera/GPS/LED rig around a Raspberry Pi 4, Arducam 64MP and GPS HAT with reusing a rooted second-hand Redmi 13C or Huawei Mate 30 that already packs a camera, GPS, storage, battery and mobile data, talking to a microcontroller over USB-C.
3 min · 634 words
Chasing an audio artifact on power loss
Subalpine Circuits' Drew tracks down the click a PCM5102A DAC made on the SA-01 synth's line out when power was lost, measuring the artifact and working through the DAC's mute behaviour and power-rail fall time to achieve silent power loss.
4 min · 870 words
Talking Zen 6 with Mike Clark at AMD In-Depth
Chips and Cheese's George Cozma interviews AMD Zen architect Mike Clark about Zen 6, the evolution from Zen 5, the differences between Zen 6 and the denser Zen 6c cores, and the design trade-offs behind them.
9 min · 1,994 words
Extending LLDB to Debug Applications That Span CPUs and GPUs
Meta engineers describe a vendor-neutral accelerator plugin architecture in lldb-server that lets LLDB debug applications spanning CPUs and GPUs, representing each as a separate debugger process in one session, with shared infrastructure upstreamed and hardware specifics left to vendor plugins.
6 min · 1,397 words
Testing Claude on real-world electronics design
Anthropic and Diode Computers asked Claude to design three electronic devices, a speakerphone, a voice-assistant hub and a handheld virtual pet, from start to finish. The prototypes worked without any fixes to their circuitry.
13 min · 3,002 words
Infineon says it has processed 300-millimeter silicon carbide wafers into functional power devices at its Villach fab using existing silicon equipment, a first in a full-scale fab, offering 2.3x more chips per wafer than 200 mm; the wafer debuts at electronica 2026.
3 min · 625 words
EJ Fox on building the touchscreen 'Big Board' that Steve Kornacki used live on NBC and MSNBC on election night 2018: from a January prototype to a broadcast tool, the stress of software performed live on air, and the couch-to-TV deploy workflow.
1 min · 340 words
Scriptable GeoDNS: Leaving AWS for Bunny
To optimize the performance metrics of this blog, I relied on *Amazon CloudFront* to serve the assets,<sup>1</sup> and *Route 53* for the geo-aware DNS<sup>2</sup> to direct readers to the closest host serving the content. But Route 53 is expensive: US$0.50 per zone per month, US$0.70 per million geolocation queries, and US$0.75 per healthcheck per month[1], plus US$1 per month for the KMS key[2] needed to sign zones with DNSSEC. When I discovered that Bunny DNS is almost free[3] for up to 500 domains and features scriptable answers, I made the jump! Plus, it’s a European…
7 min · 1,621 words
Newsletter 007: You Can Just Do Things
It’s already the 5th of the month, and we’re back at it again. We’re writing to you from the Every Thesis conference at Pioneer Works. We’re having conversations with artists, technologists, founders, and operators about what great human work looks like after this wave of automation. Josh Miller, the founder of The Browser Company, began with a provocation: what we need are new ideas, and those new ideas exist within us. AI can help us express our ideas more fully, but won’t be the engine of new ideas itself.
5 min · 1,234 words
Inside GitHub's new Git storage architecture
GitHub details a new Git storage architecture built so thousands of agents can push to one repository at once, and how the redesign works under load.
15 min · 3,392 words
Docker Sandboxes vs. Containers: Key Differences
Docker explains how Sandboxes differ from containers: microVMs with their own kernel and private Docker Engine, and when to use each for isolating AI agents.
13 min · 3,098 words
How we continuously validate formal specifications in production
Datadog engineers describe FETCh, a system that continuously checks production behavior against TLA+ specs and caught a subtle bug conventional testing missed.
13 min · 2,976 words
Starfall – Affordable, routine access to microgravity
SpaceX introduces Starfall for routine microgravity access on Starship: up to 1,000 kg payloads, 1 kW power, continuous Starlink connectivity, aimed at biotech, pharma, and materials work.
1 min · 239 words
Quasi-logarithmic kitchen timer
lcamtuf builds a kitchen timer with an old-school analog panel meter on a quasi-logarithmic scale and a knurled adjustment knob, walking through the UI design, an 8-bit AVR DB microcontroller driving the meter with PWM, an AD8531 op-amp speaker driver, the rotary encoder and parts list, with source code linked.
3 min · 785 words
How AI agents run insurance's back office
Harper Engineering describes ambient agents that continuously handle insurance back-office work—reacting to events, drafting actions, and keeping humans in the loop for 6,000+ small businesses.
7 min · 1,508 words
The Ultimate MoE Training HandbookHow to train a Mixture-of-Experts model at scale, setting by setting, with the runs that back each choice
Hugging Face publishes a free, hands-on MoE training handbook distilled from hundreds of thousands of GPU hours: expert sizing, routers, load balancing, all-to-all communication, upcycling, and serving.
185 min · 42,497 words
Inside Replit's New Sandbox Engine, Part 1: The Path to MicroVMs
How Replit rebuilt its sandbox engine with microVMs for stronger isolation, bottomless storage, and fast startup across hundreds of millions of projects.
10 min · 2,275 words