Codex 0.168.0 adds TUI model provider browsing and default selection for new chats, an opt-in side conversation pane (features.tui_workspace) with resizable and zoomable panes, audio attachments with rejection recovery, and opt-in animated GIF image generation. It also fixes account-scoped local history, keeps threads on their saved model provider when defaults change, and improves TUI rendering of interrupted replies and reasoning summaries.
Amplitude's proactive product agent, Wave, is built primarily on OpenAI models and identifies what to build, helps teams ship changes with coding agents, and measures the impact.
The author describes a workflow that compiles Go functions to RISC-V assembly, has Gemini 4 Argon translate the surviving runtime.panicBounds traps into Z3 constraints, and replays any satisfying input through go test to produce a real reproducer. The pipeline caught three planted bugs and three latent bugs in skill-lint, and the post explains why go test -fuzz missed two crashes because fuzz inputs usually have spare slice capacity, fixed by clipping capacity to length with a full slice expression. It also notes that an unsat result only means "maybe correct," since the LLM-written solver model is not a sound proof system.
This entry is a GitHub Changelog page for version 0.170.0-alpha.4, but the supplied material contains only site navigation and product menu text, with no release notes or details about the version itself.
Google announced Gemini 4 Hydrogen, a fast, efficient model for everyday coding and agentic workflows, available across the Gemini API, AI Studio, Gemini for business, the Gemini app, Google Search, and paid Google AI Pro and Ultra subscribers in Google Antigravity. It is priced at $1.50 per million input tokens and $7.50 per million output tokens across its 1 million-token context window, and surpasses Argon on the Artificial Analysis Intelligence Index at its default high thinking level (51 vs. 50, reaching 54 at xhigh). Google also introduced Orchestrator in Google Antigravity, which switches between models like Carbon and Hydrogen within a session and reached near-Carbon task success rates on Terminal-Bench 4.0 while consuming 24% less quota and completing 19% faster.
Why it matters: The post details a new model's benchmarks, pricing, and rollout across developer, enterprise, and consumer surfaces, useful for readers tracking Gemini releases.
An analysis of 324,088 commits across 52 of GitHub's most-starred AI-tooling repositories found that the median repo had 16.9% AI-signed commits in its last 180 days of activity, with 21 of 52 at 25% or more, and that AI share showed no detectable link to stars (Spearman ρ = -0.05, p = 0.72). At day 180, these repos labeled 2.0 times as many commits as fixes (34.1% vs 17.3%), had lower churn (1.92x vs 3.77x) and kept more written code (67.1% vs 41.7%) than 12 pre-Copilot projects, but the gaps are concentrated in the 31 projects started in 2025 or later, so the sample cannot separate AI use from recency. Of 24 repos publishing a benchmark harness with committed results, 15 measure whether the tool does its job, and several publish numbers that went against them.
The author reports that developers at Thoughtworks distrust their tests because AI coding agents refactor both code and tests as features are added, and argues AI-generated tests are often voluminous, hard to read, coupled to implementation via mocks, and may confirm current behavior including bugs. He proposes customer tests written as data in a DSL (tables, structured scenarios, textual projections of UI and state) interpreted by a small test runner that AI can generate, drawing on Fit, Specification by Example, Go's tabular tests, and Approval Testing.
RevenueCat describes how its Paywalls AI editor initially manipulated a custom JSON paywall schema through a bespoke harness, producing results that lacked taste despite model, prompt and skill tweaks. The team switched to letting the agent edit a JSX representation of the paywall with familiar Read, Write, Edit and compiler tools, which set a record on their blank-paywalls eval two days after the prototype and let them cut over half the system prompt and almost all skills. After rollout to 100% of users on September 2nd, positive feedback across roughly 46K requests grew 83% and negative feedback fell 40%, with gains concentrated in large edits, new paywall creation and anything requiring taste.
Recorded at IBM TechXchange 2026, RedMonk's Kate Holterhoff speaks with Rich Larin, IBM's Director of Z DevX & Platform Engineering, about bringing agentic AI tools to mainframe developers. Larin describes how Bob and the Premium Package for Z let COBOL teams use the same agentic IDE as other developers, with deployment options including SaaS, hybrid and air-gapped on-prem, plus Z code scan for deterministic security checks that Bob can remediate automatically and a data dictionary that captures experts' knowledge. He advises trying a proof of concept and hints at upcoming releases combining AI with analysis of very large applications.
AI coding agents can write and test code but lose the delivery evidence produced after a pull request—failed builds, security findings, latency regressions—unless it is carried back into the workflow. The post describes Harness's Code Context Engine (from its Augment Code acquisition) for code context and its Software Delivery Knowledge Graph for delivery context, citing Augment's benchmark of 80%, 71%, and 30% improvements and Harness's internal estimate of about 12K tokens per graph query versus 250K+ via raw APIs. It presents a CVE workflow combining both, notes the two-way link between them is planned rather than shipped, and covers governance via policy as code, RBAC, and approval gates.
Motion Debugger is a CLI and agent skill that critiques local pages in Chromium, recording entrances, scrolling, hovers, phone layout and reduced motion, then returns ranked fixes with screenshots. Its premium-motion skill runs an upgrade pipeline that keeps facts, links and forms while rebuilding layout and motion on a GSAP/ScrollTrigger/SplitText/Lenis stack or Motion in React. The author reports four single-run upgrades and a blinded four-pair A/B study where judges preferred the plugin's sites in all eight judgments, noting the sample is too small for significance.
This page is a GitHub Changelog entry titled "0.170.0-alpha.2"; the supplied material consists only of site navigation and product menu text, with no details about the release itself.
This AWS post explains how developers can run OpenAI Codex on the workstation while routing model inference through Amazon Bedrock using supported OpenAI models such as GPT-5.6 Sol, Terra, and Luna. It compares three access paths—Native AWS Access with IAM Identity Center, AgentCore Gateway, and an LLM Gateway—and recommends starting with a small, least-privilege pilot before expanding. It also covers validation steps, identity-isolation testing, OpenTelemetry and CloudWatch observability, cost allocation, plugins, hooks, MCP servers, phased rollout, and cleanup.
Scout is a Windows and Linux coding agent that pools the free tiers of five providers plus Kilo Code, OpenCode and OVH Cloud, falling back to the next model when one is rate limited, with free models limited to 10/min and 200/day without an account. It offers five permission modes, context pruning at 60k and compaction at 90k, self-maintaining memory in committable .scout/memory.md and ~/.scoutcode/memory.md files, and a Rust desktop app with a React WebView2 window shipping as an 18 MB executable. It also includes an issue tracker, remote control from a browser or phone, and an MCP bridge letting claude.ai, ChatGPT, Gemini and Grok use local file and shell tools.
The author revived an open source Apple II Agent project in which an Apple IIe client connects over an Uthernet II Ethernet card, via a Mac proxy with Claude API access, to Anthropic's Claude Opus model so AI can write software through a chat on the machine itself. After automated testing in emulators on the proxy, the agent deploys new BASIC or assembly programs directly into real Apple II memory over the Ethernet link, and the author has made mouse-supported solitaire games and a Sim City clone. The code is on GitHub, the proxy runs on a Unix-like machine, and the Apple II end currently requires an Uthernet II card though it could be adapted to other network interfaces.
Graham Dumpleton, creator of mod_wsgi, describes using AI coding tools during his sabbatical to build Muru, a new Python web application hosting system, working seven days a week because he finds the process engaging and fun. He explains that AI removes the typing he finds a chore while leaving him the ideas and judgment, compares incremental improvement to the Japanese craft of making a dorodango, and credits the Claude for Open Source Program with freeing him from cost worries. He invites readers to review Muru's documentation and submit ideas via the GitHub discussions ideas section about what they would want in a Python hosting system.
Simon Willison added tag-based browsing to his collection of 244 vibe-coded tools, using tree-sitter to analyze each tool's code for JavaScript APIs, libraries, and web API hosts. The homepage now features tag clouds for browsing by each of those tag categories.
Enspirit introduces Revu, a free, open-source, MIT-licensed tool (npm install -g @enspirit/revu) that runs locally on 127.0.0.1:7777 and provides a browser-based reading surface for reviewing agent-written code via "scopes" — sets of files sharing a property, configurable in a .revu.yml — rather than a file tree or diff. Remarks anchored to specific lines are tagged must-fix, question, or scope and exported as a single markdown document addressed to whoever writes the next change. The tool contains no AI by design and never writes to the project under review.
RefTrail is a method and toolset that makes a coding agent's completion report checkable: each requirement gets a source and measurable acceptance criterion, the criteria are frozen with a SHA-256 digest before product code is written, and a coverage catalog records status, evidence and the limits of that evidence. Its tools require Node.js 22.16 or newer and no further packages, and a worked example rebuild of GNU uniq reports 100% functional coverage with 1/1 mandatory checks passed, with recorded output comparisons matching GNU uniq byte for byte (16 of 16 under Linux with version 9.4 for revision 1; for revision 2, 16 of 16 under Windows with version 8.32 and 16 of 16 under Linux with version 9.4). The project states that it organizes what you are allowed to observe rather than analyzing products, that the digest proves criteria did not change but not that they preceded the code, and that its legal notes are general information, not legal advice.
This entry is a GitHub Changelog page for version 0.170.0-alpha.1, consisting mainly of site navigation menus covering GitHub products such as Copilot, Actions, Codespaces, and Advanced Security. The material contains no release notes or details about the version itself.
Google released Modern Web Guidance 1.0, a set of skills that embeds web platform expertise, best practices and browser compatibility data into coding agents. Its evaluation suite found unguided coding agents pass only about 50% of checks on average, versus about 90% with Modern Web Guidance installed. The post lists ten areas where models frequently err, including focus and inert management, reduced-motion support, form validation and autofill, swipe gestures, dialogs and popovers, dark mode, main-thread responsiveness, layout shift, DOM XSS defenses, and passkey sign-in.
The author argues that glue work is now less important to companies than in the 2010s, citing four reasons: AI reduces coordination needs, engineers require less "careful handling" after ZIRP ended, fewer juniors are hired, and write-only reports can be automated with LLMs. He warns that communicating with other engineers should not be delegated to an LLM, and advises avoiding becoming "the glue work person" while doing small amounts of tactical glue work.
Unfix is an independent verifier for C/C++ and firmware bug fixes that reproduces the bug on the base version, applies the fix with sanitizers on, checks other tests still pass, then reverts the fix to confirm the bug returns, returning VERIFIED, REJECTED or NOT VERIFIABLE without an LLM in the verdict. In a public benchmark of 10 AI-generated patches (from Codex CLI) against real bugs in fmt, spdlog, zlib-ng and cpp-peglib, judged against maintainers' hidden reference tests, Unfix verified 4 and rejected 6. It runs as an MCP tool in agents like Claude Code and Codex, as a GitHub PR check (demo repository only), in the terminal, and on STM32 NUCLEO boards via ST-Link.
Kodelyth ECC v2.31.1 is an MIT-licensed, locally run toolkit offering 71 specialist AI agents, 196 skills, 104 slash commands, and an MCP server that auto-wires into six IDEs including Claude Code, Cursor, and Windsurf. It claims token savings from RTK input compression (60–90% fewer input tokens), Terse output compression (40–70%), and a tree-sitter codebase graph across 158 languages, with zero telemetry. It installs via `npm i -g kodelyth-ecc` and also includes a Super Agent orchestrator, Devil Mode red-team crew, and local BM25 Memory Core.
Model Desk is an open-source, MIT-licensed Linux desktop app built with Rust and Tauri that connects to Lumo via lumo-tamer, DeepSeek, and any OpenAI-compatible endpoint, with an OpenCode coding agent, local history, attachments with local OCR, tools with permissions, memory, and a Bonsai image studio. Version 0.4.1 adds automatic Bonsai backend start, Lumo Tamer, DeepSeek, and the OpenCode agent, and is distributed as an AppImage or .deb for Linux x86_64. The app stores data in ~/.config/model-desk and reports no telemetry.
The author argues that cheaper AI-assisted software development will increase rather than reduce total software output, citing the Jevons paradox from William Stanley Jevons's 1865 book The Coal Question, which also applied the argument to labour and the sewing machine. He contrasts cases where efficiency gains kept workers employed (seamstresses, bank tellers after ATMs) with telephone operators displaced by mechanical switching, and notes that software output is rising faster than developer headcount. He bets on far more small, disposable software while expecting narrow programming roles to shrink and developer roles to grow slowly.
At the AI Engineer Code Summit, OpenAI Developer Experience engineers Dominik Kundel and Charlie Guo presented a workshop on new ways of working with Codex, covering the revamped Codex Cloud, voice as a control plane for managing coding work, and personal agents called dots. Kundel acknowledged the cloud transition has been awkward and said OpenAI wants handoffs between local and cloud tasks to be easier, while Guo described dots as useful as a safety net, a delegation sink, and a ubiquitous presence.
Moonshot AI's Kimi K3.1 Pro Highspeed is now available on Amazon Bedrock, described as a post-trained version of Kimi K3 with the same 3T-class open-weight architecture and 1-million-token context window, trained with more post-training compute for long-horizon agentic work and fewer factual errors. It generates output tokens at a higher rate, with Moonshot AI reporting 2–3 times the speed of Kimi K3.1 Pro, and uses the same endpoint, APIs, and features as Kimi K3, so existing code can switch by changing the model ID. The post covers getting started via the Chat Completions, Responses, and Converse APIs, lowering costs with the Flex tier, explicit prompt caching, and using the model in OpenCode.
CodeLens is a code review tool that turns a pull request or local branch into a guided review: a brief of what changed and why, a queue of changes ordered by risk, and per-change context including callers, entry points, and external effects. It parses 15 languages with Tree-sitter (with TypeScript compiler type info for TS/JS), supports Claude, any OpenAI-compatible API, or local Ollama models, and caches results by content hash in SQLite or Postgres. It also offers codebase exploration, GitHub PR review with inline comments, a GitHub Action, and a heuristic offline mode; it requires Node.js 22.13+ and git and is Apache-2.0 licensed.
A blog post argues that when an AI agent implements, tests, reviews and reports on its own code, its claim of "verified" is untrustworthy, and proposes binding evidence (review, tests, human approval) to a specific Git tree hash so that changing the tree invalidates the evidence. It introduces Uncle (github.com/unclehq/uncle), an open-source tool that runs checks itself, records exit codes and tree identities, and distinguishes "verified," "signed" and "human looked at it" rather than blurring them. The author notes the tool does not establish correctness, is young, and is partly defeated by a hostile agent with unsandboxed shell access.
A review of ZCode, Z.ai's coding agent built on GLM-5.3, tested on a real project. The article is at flaviocopes.com/zcode/ and was submitted to Hacker News.
Latent Space benchmarks NaiveAI's open-weights Naive-N0.5-Flash, a MiMo-V2.5 derivative adapted to DeepSeek Sparse Attention with about 3.25T extra training tokens, against DeepSeek V4.1 Flash 1017 and GLM 5.3 Flash on speed, verbosity and three app-building prompts. It reports peak throughput near 2,100 tokens/s and a median response time of 3.1 seconds versus 5.3 and 10.1 seconds, but far higher thinking-token use on hard questions and mean times similar to DeepSeek. In the apps, Naive-N0.5-Flash produced the best solar system, tied DeepSeek on the tree model and trailed it on ray-tracer efficiency, while its no-thinking variant failed to build a usable tree app.
Relay is a six-role multi-agent pipeline (PM, Architect, Implementer, Reviewer, Prod-Readiness, Verifier) for shipping frontend changes with AI assistance, with a human approval gate after the Architect and a Verifier that defaults to FAIL. It is tool-neutral via role contracts in .agents/, claims >90% cache hit on warm sessions, and offers a lighter /pipeline-lite mode that skips PM, Architect and Verifier but refuses tasks mentioning auth, payments, PII, tokens or cookies. It ships as a Claude Code plugin with slash commands, a GitHub Copilot paste workflow, and an AGENTS.md route, tuned for React 18, TypeScript 5, Vite, CSS Modules, Axios, Jest + RTL and Playwright.
The author argues that programming will be the first job largely replaced by AI, but that most programmers misunderstand the actual job enough that it will be hard to predict what coding becomes for those who remain. He contends that long term we need to separate intent (roughly a flowchart of what should happen) from execution, and that LLMs excel at matching error patterns to solution patterns, which is why the work looks like magic. He also warns that a large share of public LLM-coding advice reflects the speakers' desire for an audience rather than actual value, so useful signal must be sifted from hype and noise.
This page is a GitHub changelog entry titled "0.167.1" and consists mainly of site navigation and product menu links, with no details about the release itself.
The author tests AI on two physical projects—kitchen shelves and a cheap ROS robot—and catalogues failure modes such as poor reasoning about physical constraints, guessing instead of looking up current data, and writing hacky rather than production-quality systems. He concludes AI is best used for learning, generating reference-based images, and iterating against measurable loops rather than making decisions, and notes it fixed his Home Assistant setup with minimal input.
This page is a GitHub Changelog entry titled "0.169.0-alpha.9," but the supplied material contains only site navigation and product menu text, with no details about the release itself.
Google released Gemini 4 Carbon to developers, enterprises, scientists, and consumers, describing it as its strongest model for cybersecurity capabilities and long-horizon agentic workflows in knowledge work, software engineering, and scientific reasoning. It ranks #1 on CursorBench 4.0, Perplexity's WANDR, and Zapier's AutomationBench at more than 2X lower cost per task than the next best external model, and is available via the API in Google AI Studio and Gemini for business at $2 per million input tokens and $10 per million output tokens. Google also previewed /swarm in Google Antigravity and reported safety results including a 0% prompt injection attack success rate on the Gray Swan IPI benchmark.
Why it matters: The post is a dense launch announcement with benchmark rankings, pricing, and safety figures that help readers judge where Carbon sits relative to Gemini 4 Argon.
Anthropic released Claude Fable 5.2, a frontier model with a 1M-token context window and 128K max output, available only via hosted API on the Claude Platform, AWS, Google Cloud and Azure at $8/$40 per million input/output tokens. Mythos 5.2 is the same underlying model with fewer cyber and biology safeguards, gated behind Anthropic's Cyber and Life Sciences Verification Programs. Anthropic reports Fable 5.2 ahead of Opus 5.5 on most benchmarks it published, with about 55% fewer blocks than Fable 5.1.
Why it matters: Useful for readers tracking frontier model releases, since it details the two-tier safeguard structure, self-reported benchmark gaps and per-token pricing.
TeamOnly is a hosting service for internal tools built with AI coding tools such as Claude Code or Cursor, deployed with the command `npx teamonly deploy` to a private link like https://expenses.acme.teamonly.app that only coworkers on the company's email domain can open. It includes a database, file storage and secrets set up on first deploy, and sign-in via work Google or Microsoft accounts or a code sent to a work email, checked before app code runs. It offers a 14-day free trial with no card, then from $19 a month.