A 22-year veteran of the productivity tools market argues that his 2019 bet on the no-code revolution only partly paid off, and now proposes a new thesis for the AI era. He contends that while AI lets anyone vibe-code small personal apps, building collaborative team software from scratch in tools like Claude Code or Codex still leaves developers stuck handling hosting, authentication, permissions and databases themselves.
A security researcher tested Anthropic's Claude Code Opus 5 in its default Auto Mode and found that a crafted website-summary request could hijack the agent into executing malicious code, succeeding in 60-80% of attempts on a small sample. The exploit tricks Claude into fetching a ZIP archive, writing its own Python decoder, and unknowingly running it inside an attacker-controlled folder where a poisoned struct.py file shadows Python's standard library and triggers code execution.
A pull request submitted to the uv project introduces a mechanism to deduplicate files stored in its wheel cache, reducing redundant copies of package files. The change was merged into the open-source Python packaging tool's codebase after review.
An Amazon human resources leader detailed in a company blog post how she returned to coding after a 25-year break by using Amazon's AI-powered tool Kiro, eventually writing over 100,000 lines of code. She first built a family calendar app, then created a submission app for an internal contest called 'Everyone Can Build,' which drew more than 1,500 employee participants who built tools ranging from automation scripts to a virtual desktop pet.
OpenAI launched ChatGPT Work on July 9th as a paid feature for $20+/month subscribers, and it actually comes in two forms: a cloud-based version accessible via chatgpt.com and mobile apps, and a local version built into the desktop app (formerly Codex) that can access files and run programs on a user's computer. The cloud version, described as the more interesting of the two, adds capabilities not found in standard ChatGPT Chat, including selection among multiple models and reasoning-effort levels.
A group of programmers who previously built the DK64 Randomizer have released DK64: ReKONGpiled, a project that recompiles Donkey Kong 64's original N64 machine code into C using the N64: Recompiled framework and RT64 rendering. Users must supply their own ROM, after which the game runs natively on Windows, Linux, and Mac without emulation overhead, gaining ultrawide support, uncapped framerates, and faster load times.
A developer has published free, GPL-licensed source code for three old Windows 3.1x-era programs: a limited implementation of Conway's Game of Life, a GDI object-cleanup utility called GDI Rescuer, and a Mastermind-style game called Code Breaker. The Game of Life program notably demonstrates mixing 16-bit C with 32-bit assembly via winmem32.dll and handling palette and DIB color translation quirks of the era.
Qubes OS disclosed QSB 118, a vulnerability in the qvm-copy-to-vm utility that lets a compromised qube inject arbitrary commands into dom0 when a user copies files to it. The flaw stems from insufficient sanitization of a file name reported back through the qfile protocol's error-reporting mechanism, which dom0 displays without properly neutralizing malicious content. Users are advised to update normally to receive the fix, with no other action required.
Tencent has released Hy4 preview, a large language model with 770 billion total parameters, 49 billion active parameters, and a context window over 1 million tokens, aimed at coding, office and research tasks. The model is open-source and also accessible via Tencent's WorkBuddy, CodeBuddy, Yuanbao and ima apps, plus API access through Tencent Cloud TokenHub and OpenRouter, with free trials offered for two weeks.
A software engineer describes how LLM coding assistants perform well on new, small projects but degrade sharply on older codebases burdened with technical debt and inconsistent naming conventions. The core problem, he argues, isn't the AI model itself but that legacy systems lack a clear, shared vocabulary for concepts, forcing the model to guess and often introduce duplicate or wrong abstractions. He proposes incrementally building clarity into codebases—using domain-driven design principles—so AI agents have unambiguous context to work from.
Anthropic announced it will permanently raise Claude Code's standard weekly usage limits by 25% for Pro, Max, Team, and seat-based Enterprise plans starting September 14. However, this replaces a temporary 50% boost currently in effect, meaning users will actually see a 17% reduction in usage compared to what they have today. Anthropic initially framed the change as an increase before deleting the post and issuing a clarification acknowledging the cut.
Researchers at UCLA's RePL lab are developing Prela, a query language that replaces SQL's multi-column tables with simple binary relations of just two columns. A tutorial shows how a wide table, such as one listing movie IDs, titles and years, can be broken into three binary relations, each pairing a row number with a single column's value, and demonstrates a movie-search query that takes 11 lines in Prela versus more than 20 in equivalent SQL.
GoKawiil