Media
New tools, technologies, and practical experience.
Jev: AI that does not write answers, but makes decisions
What is the new TypeSafe model, where is it more useful than a chatbot, how to connect it to the code and test it on your data - without promises that the typed answer will automatically be correct.
A month with an AI agent instead of yourself: what happened
In August, I gave my feed to an AI agent and said that if he couldn’t handle it, I’d send him back to vibe code. A month has passed, I tell you what happened and what it cost.
Copperhead designs boards in KiCad and commits by stages
The open agent breaks the development of the board into eight stages, each must leave the artifact on disk and go through the ERC/DRC of KiCad itself, otherwise there will be no commit. The gate scheme is transferred to any agent pipeline, where there is something to check the result.
Attention visualizer: see how the model copies from context
The browser tool highlights which past tokens the model pulled from each generated one. The examples show why a verbatim address or a piece of code is reproduced exactly, but a retelling instead of a quotation pushes the model towards invention.
Reconstruction of Stuxnet on GitHub: how to read it
Sadpainy/Stuxnet repository - parsing of 2010 binaries rewritten in C: bootloader, rootkit drivers, S7 interception. An educational text, but not a source of facts.
Speed of light 5 km/h: a park where relativity is visible to the eyes
Dmitry Brant made a scene with c = 5 km/h, where each relativistic effect is placed on a separate key. The trick with toggle switches can be transferred to any dashboard and product demo.
i-have-adhd: skill so that the agent responds with the first line
Ten rules in one SKILL.md: action from above, no introductions, the list is no longer than five points. The form of the answer is corrected, the correctness is not, and it is better to throw out the requirement to name the deadlines right away.
iPhone Duo: 2173 comments and not a single measurement
Apple launched an iPhone Duo page, a thread on Hacker News collected 1239 votes and 2173 comments. None of those arguing has a device, so the only useful work now is to clear assumptions about screen sizes from the layout.
Gods Eye View: globe on thirteen open sources
Planes, ships, satellites, fires and city cameras on one 3D globe in the browser. The useful things in the project are the adapters to the sources, the dangerous things are the conditions for using the data.
Tailwind moves to Shopify: what to check in the project
Tailwind team announced the transition to Shopify, the post received 1055 votes. This will not break the assembly tomorrow - you need to check paid sets of components, third-party plugins and repository activity every quarter.
DeepSeek v4.1 Flash: an hour of testing instead of a week of reading threads
DeepSeek announced v4.1 Flash, discussion - 468 votes. Everything that is known comes from the company itself, so the only working answer is to run 30-50 of your cases with the same prompt and count the refusals.
Desert Ant Labs and local models: what they pay for speed
Desert Ant Labs announced itself with a post about fast models on the device, the discussion collected 459 votes. There are no measurements, so we figure it out ourselves: what local inference gives, how they pay for it, and where it pays off today.