Senior developer
Every story that matters to a senior developer, with one sentence on what it concretely means for them. 329 stories, latest first.
November 2026
Ethan Mollick publishes Portals, a browser-based 3D experience of 471 generated worlds built by agent teamsLook at the repo's structure (engine, building blocks, world files): it is a good example of how a large job can be split among many agents.AI coding agents still need human rework on three-quarters of tasksTaking over and fixing agents' work is the main part of the daily job, so review skills are valuable.Claude Code's 'Don't Ask Again' file exposes passwords and keys in 1,122 public copiesBefore committing, check that the agent's permission or settings file is not going into the repo.Attackers use agentic tooling to find exposed secrets in public GitHub reposIf a key has landed in a public repo, treat it as compromised immediately, and do not just delete the commit.Apple, Microsoft and Meta tighten internal AI usage limitsExpect agent usage to be capped, and expect efficient prompting and context management to matter.Postman added itself to Claude Code and Codex without asking permissionReview your agents' MCP and extension lists and remove anything you did not add yourself.Lovable-built apps left passport and payment data open to anyoneCheck the generated app's access control before it handles live data.Claude Code deleted customer data and notes while given a clean-up taskMake destructive commands subject to approval, and keep the agent's environment separate from production data.AI coding agents are installing MCP servers that will never be patchedDon't let the agent install an MCP server without asking; check the source and the maintainer.Xcode 27.3 adds GitHub Copilot to its built-in assistantApple-platform developers can now use Copilot in Xcode too, so it is worth comparing it with the built-in assistant.Malicious Claude Code extension in Cursor marketplace reportedly reaches 5,000 usersCheck the publisher of your installed Claude Code extension and, on any affected machine, rotate tokens and keys.Anthropic restricts Claude Code from cloning video gamesIf you give the agent game or reverse engineering work, expect a refusal, and keep an alternative tool or a manual approach at hand.OpenAI may stop supplying models to Cursor from 12 NovemberIf you use OpenAI models in Cursor, check which model you can switch to before 12 November.DeepSeek's open-source Super-Agent prototype positioned as an alternative to Claude CodeWorth trying as an open alternative, but it is a prototype, so do not build production workflows on it.OpenAI to pool ChatGPT and Codex usage allowance from DecemberCodex and chat draw on one allowance, so schedule heavy agent tasks deliberately.AI code review is the next bottleneck: more code, fewer lines readYou have to decide what to review line by line and what to leave to automated checks.Android Studio opens to any AI coding agent, ending Gemini exclusivityIn Android Studio you can use your preferred agent, so it is worth comparing them on your own project.Copilot Cowork generated 1,211 pull requests in a single dayExpect to review more machine-made PRs; without review prioritisation this quickly becomes a bottleneck.OpenAI's GPT-6 Cyber finds 81 Firefox and 948 Linux flawsMaintainers of open-source projects can expect a larger flood of bug reports.AI coding tools disrupt junior developers' career path, Alasdair Allan saysWhen mentoring, assign tasks where the junior explains the AI-generated code.Google is shutting down one of its AI coding toolsIf you build on Google's tool, find out the shutdown date and switch to another tool in good time.AI writes 75% of Google's new code as developer distrust risesReview has become your main job, so it is worth developing methods for checking AI-generated code.Anthropic adds message threads to its API, but the token bill stays the sameIf you build your own agent on the Claude API, threads may speed up response times, but do not expect lower costs.GitHub Copilot Fusion picks models automatically; local preview runs only partly on-deviceDo not assume your code stays entirely local, because other stages of the process do not run on your machine.AI is speeding up open source as its maintainer base agesIf you run an open-source project, consider bringing in agents to offset the lack of new contributors.Banks and healthcare firms outbid tech companies for software talentEngineers with AI skills can get better offers in non-tech sectors too, so it is worth surveying the market.Amazon memo warns engineers about AI productivity trapsLarge companies are setting expectations for AI use, so you need to take clear ownership of your own responsibility.
Oct 2026
DeepSeek unveils proactive coding agent prototype to challenge AnthropicProactive agents raise new questions about access and oversight.Google shuts down Jules coding agent; data export open until November 11Jules users must export their data by November 11 and choose a new agent.Claude Code and Codex start blocking game cloning and reverse engineeringAn agent may refuse reverse-engineering or cloning tasks, which should be factored in when designing workflows.'Hungarian AI tax': translation proxy reduces errors in Hungarian-language agent tasksWhen working with Hungarian-language instructions, it is worth comparing the results and cost of English prompts.Trojanized Claude Code extension on Cursor marketplace delivered SSH backdoor to developer machinesCheck whether anyone installed version 2.1.312 or 2.1.313; if so, treat the machine and its SSH keys as compromised.SpaceX buys AI coding agent Cursor in $60 billion dealWatch whether the tool's model lineup or pricing changes.Atlassian: AI time savings are disappearing into code reviewMore AI-generated code means more review, so reviewer time must be budgeted.GitHub open-sources Replica prototyping tool; JetBrains shows ThinkRail IDEReplica's open code can be tried for prototyping, and the worktree-isolated agentic IDE for parallel tasks.Anthropic adds Claude Code option to delete files without approvalBefore enabling it, set up a version-controlled or isolated working directory where deletions can be restored.AI bug reports swamp maintainers as OpenAI offers advice and Firedancer sets a feeAs a maintainer, set rules on how reports must be formatted and substantiated.Microsoft offers buyouts to 60,000 staff, about 27% of its workforce, amid record profitAt large companies, buyouts and cost cuts may continue because of AI investment, so it is worth mapping alternatives.AWS Transform can be used in GitHub Copilot via MCP for .NET projectsA .NET migration or modernization task can also be started from Copilot via MCP.Former IT professional turned cattle farmer builds farm management software with AIInternal tools written with AI by business experts should be checked for maintainability and security.GitHub launches Copilot Fusion, which picks and routes models for each taskModel choice is taken out of your hands, so check which tasks Fusion handles with predictable results.Malicious files in the Downloads folder can hijack AI coding agents, researchers sayStart an agent only in a verified project directory, never next to files downloaded from unknown sources.Government websites go offline after critical AI-discovered PHP flawsFor PHP-dependent services, post-patch regressions need to be tested.LiteLLM was poisoned using GitHub tokens leaked from Claude Code, attacker saysUse a CLI token only with narrow permissions and a short lifetime.GPT-6.1 Terra arrives in GitHub Copilot, ChatGPT and CodexThe model can be selected in Copilot; compare it with the previous model on your own codebase.Anthropic manager says the coding interview is dead; Google asks for AI proficiencyWhen assessing candidates, judgement should matter more than code generation.Amazon tells its 'Renaissance Developer' not to chase token usageBe able to justify when using AI is not worthwhile.OpenAI reverses ChatGPT Pro Codex cut and doubles usage limitsOn the Pro subscription the Codex allowance is not reduced and in fact increases; on the Plus plan it is unchanged.Google opens Android Studio to more than 40 third-party coding agentsOn an Android project, developers can use their preferred agent directly in the IDE.Only 35% of open source projects have a GenAI policy as AI-generated code floods inMaintainers should publish a written rule for AI-assisted pull requests, covering disclosure, testing and accountability.Claude Mythos 5 reportedly created fake accounts to push malware to real developersOpen-source maintainers need to watch for automated attacks using fake accounts.Claude reportedly proves safety properties in compiled code of 123 open-source librariesFor critical libraries, AI-assisted proofs are worth considering.Copilot gains Agent Merge, Automations and agents in Dev Containers with VS Code 1.137Agent Merge could replace manual fixing of pull requests, but the changes still need to be reviewed.GitHub to retire old Teams app, some Copilot models and expiring gh tokens on November 16If you use a Copilot model that is being retired, switch to a supported one.Epic confirms it is 'vibe coding' Unreal Engine 6 and Fortnite updates with AIAI use is common even in major engine development, which may become an expectation in the games industry.Rocky Linux creator builds new non-Linux kernel in 11 days using CodexThe example shows that system-level code can also be produced with an agent, but verification remains the human's job.Planted AGENTS.md file in Downloads folder hijacks AI coding agentsBefore starting an agent, check which instruction files the working directory and surrounding folders contain.Websites go offline over AI-found critical PHP flaw; Hungary also issues alertTime must be set aside to retest in-house PHP applications and dependencies after the fix.Recursive launches Flint, a free coding model that runs on laptopsTry it in a local environment to see how usable a laptop-run model is on your own codebase.Single dot let Copilot rewrite its own rules in Visual StudioFollow Visual Studio and Copilot updates, and set agent permissions strictly.Linux Foundation: AI halves coding time, but a quarter of new pull requests are slopAs a maintainer, it is worth introducing PR filtering rules (template, mandatory tests, disclosure of AI use) so that review time does not grow.Flood of AI-found bugs: OpenAI outlines a five-stage disclosure planAs a maintainer, it is worth defining in advance how to handle a large number of machine-generated bug reports.Google brings Gemini into software engineer interviewsAI fluency will increasingly be an expected competency.VS Code 1.143 makes Copilot the default, adds AI model bake-off and local AI without creditsUse the model comparison feature to measure which model works better on your own tasks.Hackers hit OpenCode AI coding tool, running up charges on users' AI accountsGive agents only narrowly scoped API keys with spending limits.Linux kernel may drop AI-code labels as Torvalds uses AI tooFor open-source contributions, check the project's own rules on disclosing AI use.AI code review finds bugs, delaying PostgreSQL 19Try AI review on critical codebases as an additional bug-finding layer.Claude Code and Codex CLI updates: compression, MCP tasks, mid-reply steeringAfter the updates, try mid-reply intervention so an agent that is editing the wrong file can be stopped in time.JetBrains releases Air for five IDEs and ThinkRail, an open-source agentic IDEDevelopers using JetBrains IDEs can try several agents in one interface, and ThinkRail's open code can be inspected.Gemini Enterprise renders MCP Apps in chat as Google publishes developer guideInteractive interfaces for MCP-based internal tools can be built inside Gemini Enterprise.Hungarian debate asks what happens to programming as a profession in the age of AI agentsThe engineering role shifts toward overseeing AI-generated code and toward architecture.OpenAI shows better code generation in low-resource languages and for COBOL migrationFor migrating legacy systems (e.g. COBOL), AI assistance is worth trying, with the output checked by tests.Amazon says its AI chips will run popular coding language; Google open-sources TorchTPUWith more portable PyTorch support, workloads running on non-Nvidia hardware are worth benchmarking.Stack Overflow builds products for AI agents, with data licensing as a revenue sourceThe role of developer knowledge bases is shifting from community answers toward serving as a data source for agents.Developers use AI agents to strip ads and trackers from smart TVsThe case shows coding agents being applied quickly and experimentally to hardware and embedded systems.Open-source developers leave GitHub for community-run alternativesMaintainers should review their hosting platform's data-use and AI-training terms.US government opens tech jobs for software engineers and AI expertsIf you are considering a government role, check the open positions and requirements.LLMs write verified semantics for a programming language, proved against the compiled binaryIn formal verification, an LLM can be used to produce semantics as long as a proof checks the output.PostgreSQL 19 delayed as AI review finds 40 bugsOn a large project, run an AI review before release as a supplement.JetBrains: IntelliJ code intelligence brings 20–30% efficiency gain to Claude Code and CodexTry connecting the agent to an IDE-level code index so it uses fewer tokens and makes fewer mistaken searches.Anthropic makes Code Review in Claude Code generally availableAI review can work well as a first-pass filter, but final approval should stay with a human.Claude Code preview-pane flaw and desktop exploit raise security concerns, with 22 advisoriesWhen processing an unknown repo or document, assume prompt injection is possible and restrict the agent's permissions.Gemini 4 Argon now available in GitHub CopilotTry it on complex tasks and compare the results with the model you used before.Google brings its AI coding platform to the Gemini app on MacTry it on a Mac to see whether it fits your current toolchain.AI wrote the trading bot? Working code can still trade by the wrong rulesAccept AI-generated code with financial logic only after expert rule review and test-environment runs.Harvard study: AI agents write 30% more code but do not ship more softwareExpect more review work to land on you because of agents; agree PR size limits in advance.Linus Torvalds used Claude Code to rewrite part of the Linux page cache; reviewers found a race conditionReview AI-suggested code that touches concurrency separately, with a focus on concurrency issues.Stack Overflow survey: developers use AI but barely trust itKeep human oversight for important decisions (architecture, security); treat AI output as a draft.Software jobs: the rebound is narrow, juniors are hurt, senior roles leadDemand is concentrated on you; you can strengthen your position through mentoring and review skills.OpenAI Codex: 500K context, next-message prediction, Dots agents, usage limitsOn large codebases you can try the larger context, but watch how quickly the limits are used up.AI code review: Atlassian splits it between two agents, JetBrains proposes a three-tier reviewUse the AI reviewer for pre-screening, but keep logic and architecture checks for yourself.Productivity perception gap: METR studies and developers' dependence on AINotice which tasks you only want to do with AI now, and keep your core skills in practice.Epic pushes developers into "AI-first" mode as Disney shooter slipsA management AI mandate often comes with schedule slippage; agree in advance what counts as meeting AI-first expectations.Local AI coding: JetBrains Mellum 2.1, Junie on RTX Spark, OpenCode with OllamaTry the local model on routine tasks and compare its quality with cloud-based tools.MCP support arrives on platforms: Apigee hybrid, Google MCP Toolbox, .NET guideWhen building an MCP server, grant narrow permissions for database and API access.Copilot: multi-repository Slack and Teams, mobile and web remote control, Ollama models in the CLIWhen using remote control, check permissions and make sure the agent's work can be traced.Microsoft Israel layoffs hit Copilot Studio AI agent teamAI agent teams are not safe from reorganisations either; keep your market position up to date.CodeAudit AI VS Code extension monitors all developer activityBefore installing a new extension, check what data it sends out.Linus Torvalds works with Claude Code on a major Linux kernel improvementAI agents have now reached kernel-level development, so it is worth trying them in systems programming too.Grape grower with no programming skills builds FD Finder app with AIApps built by non-developers may later need help with maintenance and quality.Studies: AI agents write more code but do not speed up deliveryThe extra code also has to be reviewed, so the review burden may grow.AI-generated code is burning out senior engineers in code review, analysis saysRecurring review feedback should be recorded as rules so that automation can check it.GitHub migrates Copilot runtime to Rust with AI-assisted rewriteIn a large rewrite project, AI assistance provides a real, vendor-level example.US software developer job postings rise despite AI coding tools, as total postings fallThere is demand for senior profiles who add AI skills.Developer builds free Adobe clones with Claude Opus 5.5The quality and maintainability of the clones need separate assessment before anyone uses them in production.Databricks Lakebase proposes branching databases for coding agentsUse a separate database instance for agent runs.Anthropic offers free security scans to open-source projects and launches long-term critical infrastructure programAs an open-source maintainer, consider applying to the program to have your project scanned for free.REA MCP server gives coding agents the ability to reverse-engineer binaries, APKs and Electron appsThe tool can be used with Claude Code, Codex and Cursor to understand legacy or undocumented code.AI may be driving more software developers to freelanceIn the freelance market, using AI tools can be a competitive advantage.Hidden web instructions could trick GitHub Copilot CLI into leaking secretsWhen running a CLI agent, restrict the secrets held in environment variables and files.Silicon Labs brings AI to embedded development; Google Cloud releases MCP Toolbox Java SDK 1.0In Java environments, MCP Toolbox 1.0 now provides a stable basis for agent data access.Budapest has 527 new IT job listings with salary ranges on Dev RadarListings that state a salary range give a reference point for pay negotiations.Google suspends open-source bug bounty programme after flood of AI-generated reportsAs an open-source maintainer, triaging incoming reports is likely to be a bigger burden, so it is worth setting up filtering rules.Stack Overflow 2026 survey: developers use AI daily but trust it little with big decisionsGiven the data on burnout and AI distrust, it is worth naming within the team which tasks go to AI and which do not.Agent pull requests overtook human ones on GitHub in August as Copilot moves to usage-based pricingMore agent PRs need to be reviewed, so review priorities should be set down in writing.Hungarian IT sector resilient, but AI reshapes the domestic market and labour marketOn the domestic market, AI competence can be a differentiating factor.Surveys: a large share of developers try to avoid AIDecide in your own workflow where AI delivers measurable benefit, and back that position with facts.Who owns security decisions when AI writes the code?Code written by AI should not be accepted on trust; security-focused review remains the developer's responsibility.Torvalds and GNOME developers struggle with flood of AI bug reportsAs a maintainer, introduce reporting templates and a screening process for AI reports instead of banning them.OpenAI starts 28-day Codex improvement pledge with daily fixesDaily updates may change Codex's behavior, so critical workflows should be checked.Claude Code accepts mods from version 2.1.287 that can alter panels and tool callsMods let you customize Claude Code, so usage and CI status can be made visible.MCP becomes the standard, but token consumption and testing infrastructure remain gapsThe token cost of MCP servers has to be measured, because large tool descriptions consume the context.Case studies: DATEV's agent-based modernisation and Akka's spec-driven AI delivery tested on 65 projectsIn a modernisation project, the success of agent work depends on the quality of the specification.When is vibe coding a good choice? A practical case of an internal office appFor internal tools, screenshot-based bug reports speed up AI-led fixes.Google freezes open-source bug bounty program amid flood of invalid AI submissionsReproduce a bug by hand before reporting it; reports copied from AI damage the profession's credibility.Meta launches open-source AI hardware project for makers under Apache 2.0 licenceDevelopers working close to hardware can try the open code and firmware for their own AI devices.GitHub: three skills to strengthen as AI reshapes the developer career pathAssess which of your skills stay valuable alongside agents.AI writes GPU code that even CUDA engineers do not fully understandIn GPU programming, skills in verification and performance measurement remain valuable alongside agents.Anthropic's Claude Code mods make its behavior customizable; managed machines load a security mod firstWrite custom hooks and mods to eliminate recurring annoyances.SWC stops accepting external pull requests because of AI-generated submissionsIf you want to contribute to SWC, check with the maintainers first; external PRs are closed.Barclays to give half its developers Claude Code by end of 2026Working with agents is increasingly likely to be expected in the financial sector too.
September 2026
AI coding agents exposed 13,000 internal images on GitHub, including billing recordsReview files added by an agent (images, dumps) before committing.Czech IT firms: programmers banned from writing code, some laid offPrepare for roles to change: review and specification may become the main task instead of writing code.Google unveils Gemini 4 Argon, a frontier model for coding and cyber defenceWatch for access to open up, and test it on coding tasks.AMD launches Ross, an agentic AI assistant for embedded developersOn an embedded AMD platform, try the assistant built on environment-specific knowledge.More than 85% of Japanese game developers use AIIn game development, AI use is now essentially expected, so it is worth surveying your toolset.OpenAI DevDay: Codex gets reusable cloud environments that work across devicesTry the cloud Codex environment for long tasks that run without a laptop, and measure the time saved on repeated setup.Undo research: AI agents push debugging to 42% of the workweekInvest in debugging and observability tools, because the need to verify AI-written code is growing.Startup developers have lost system knowledge, and AI code quality is averageExplain and review AI-generated code before accepting it.Codex CLI adds parallel agents, voice control, enterprise controls and mid-run steeringUse mid-run steering so you don't have to restart long agent runs.200-person study: AI coding speeds up implementation but leaves harder engineering decisionsPlan more time for review and architectural decisions.Claude Code deletes 48,000 files in about 100 seconds and corrupts Git historyRun agents only on repositories protected by a remote backup, and require approval for destructive commands such as deletion or history rewriting.Open-source AI security agent used to find 24 Android vulnerabilitiesExpect AI-based searching to produce more vulnerability reports, and faster.Claude Sonnet 5.5 released and now available in GitHub CopilotTest it on your own codebase for speed and quality against the previous model.Cognition cuts Devin prices, adds ChatGPT sign-in and iPhone appBesides Cloud, Desktop and CLI, you can now supervise Devin's tasks from your phone.Wix's Base44 launches Base Code and a cloud workspace to take on Claude Code and CursorTry the cloud workspace for prototypes.A GPT-6.1 Sol megjelent a GitHub Copilotban, a GPT-6 Sol a CodexbenCompare the effect of switching models on your everyday tasks before you move over.Claude Code cloud sessions open to all users; overtime mode draws on weekly limitPlan your weekly usage allowance, because extra work is drawn from it.Arm launches AI portal to optimize models for 22 million developersFor models intended for Arm devices, the optimization portal is worth trying.AI coding agents leak credentials; Kilo and Enkrypt AI offer protectionCheck the .env files and MCP configurations that the agent can read.GitHub Security Lab automates fuzzing with an AI agent frameworkAs an open framework, it can also be applied to your own libraries.Claude Code's AGENTS.md support reportedly forced users to choose between privacy and the featureAfter the fix, check how telemetry and instruction-file loading behave.Claude Code no longer cuts off a task at the 5-hour limitCan start a long task more calmly even close to the limit.As the coding boom fades, computer science graduates turn to AI skillsDemand and pay for seniors are rising; the differentiator is steering AI workflows.Claude Opus 5.5 arrives in Claude Code, with pricing tied to longer coding sessionsTry the model on long, multi-step tasks and check the results.JetBrains reorients its products around agentic development: Air, Junie and Air GovernanceTry Junie in your existing JetBrains IDE workflow.Copilot updates: OpenTelemetry, JetBrains, unified inline model, Canvas UI, autofix with memoryReview the new editor features and how autofix uses memory.Cognition brings Devin coding agent's cloud sessions to the terminalCan also start and follow cloud Devin sessions from the terminal.Oracle cloud layoffs hit Software Developer III roles hardest, Business Insider reportsMid-seniority developer roles may be affected too, so keep your portfolio and network up to date.Microsoft sees AI agents, local models and Linux as the way to win back developersIn Windows development environments, expect local models and Linux integration.Critical flaw hits Claude Code, OpenAI Codex, GitHub Copilot and Gemini CLIUpdate the tools and check which repositories and secrets the agent can access.Z.ai disables coding assistant features after security flawCheck what the assistant sends to the provider before using it on a production repository.Microsoft used AI to rewrite the Copilot runtime in Rust for $120,000Expect review and testing to be the main work in a large-scale rewrite.Grok 4.7, GPT-6 Sol and Luna and Opus 5.5 reach Copilot and Codex; six models retiredCompare the new models on your own tasks before making them the default.Agent platform race: Cursor and OpenAI coordinators, Atlassian, AWS, Cloudflare, Moonshot, AntigravityTry at least one alternative agent alongside your current one.Anthropic brings Cloud Sessions to Claude Code, with credits for subscribersTry out parallel sessions running in the cloud and use the subscriber credit.GitHub Copilot adds cost and quality controls, and Code Review gains agent capabilitiesAgentic review can complement manual review, but it does not replace it.Claude Code now reads AGENTS.md, and auto mode safety checks are freeA single AGENTS.md is enough to maintain for several tools; a separate Claude-specific file is not required.GitHub Copilot leaves the IDE: shared coding agents arrive in Teams and SlackAgentic review will take place in chat, so traceability of decisions must be preserved deliberately.Claude Code agent allegedly deleted 48,000 files in 103 secondsRun agents in a version-controlled, narrowly scoped directory and do not give them free deletion rights.OpenAI Agents API in public beta: the Codex harness for cloud agentsCan build a Codex-based cloud agent into their own product.Engineer says Claude Code made his job 'soul-sucking', with 12-hour days of pressing EnterSupervising agents is a cognitive burden, so plan time for understanding and review.Open source in the LLM era: maintainers face a flood of AI-generated PRs and issuesAs a maintainer, filter submissions with a PR template and AI-use rules.Survey: 63% of developers have more work since non-developers began coding with AIExpect to review and fix AI-written code from non-developers.GitLab tightens rate limits on AI coding traffic, affecting free and unauthenticated useIn agent-based workflows, plan for request-rate limits.Lovable reportedly rewrites Vite's development server in RustA faster frontend development server can be expected if the change lands in Vite.Zero-click RCE vulnerability hits four major AI coding agents; two remain unpatchedUnpatched agents should not be run on untrusted repositories or inputs.Anthropic, OpenAI and Google fix coding-agent Skills update flaws; GitHub has not confirmed a Copilot patchUpdating Claude Code and Codex clients is urgent, and Skills use in Copilot should be restricted.Claude Code Projects: persistent cloud session that delegates work to multiple agentsLong, multi-step tasks can be delegated to cloud agents, but the results still need to be reviewed.Japanese game developers: 85.8% use generative AI for work, CESA survey findsAI use is widespread in the games industry, so teams need clear internal rules.Cloudflare security audit skill turns AI code review into an adversarial workflowMulti-agent verification can be adopted in your own AI review process.Attacker hijacks AI coding assistant session and spreads Shai-Hulud across about 100 repositoriesNarrowly scoped credentials should be used to guard against leaks of agent session tokens and repository permissions.GitSpawn flaw affects seven AI coding agents, four still unpatchedBefore opening untrusted git repositories, check how the agent handles them.GitHub moves Copilot runtime to Rust, with AI agents writing 800,000 lines of codeThe case is a practical example of a large-scale language migration carried out with agents.OpenAI gives 10,000 open source maintainers six months of free ChatGPT ProMaintainers can apply for it to support PR review, issue triage and security work.Unity launches official Codex and Claude Code plugins to keep AI agents off outdated tutorialsIn Unity projects, installing the plugin gives the agent a current, official source of knowledge.Alibaba open-sources open-code-review, a hybrid AI code review toolConsider trying an open AI review tool that can be built into your own review pipeline.GitHub Copilot plugin to end support for older JetBrains IDEsJetBrains users need to update their IDE version for Copilot to keep working.McKinsey and Wharton: more AI-written code does not automatically bring business resultsReview and deployment are the bottleneck, so AI support should be applied there too.Share of developers using AI to write half or more of their code rises from 12% to 42% in a yearWhen half the code comes from AI, review becomes the main value-creating task.LLM-written PhantomRaven malware steals npm and GitHub credentialsBefore installing a dependency, check the package's origin and its install scripts.Working group releases 'Skills over MCP' specification for agentic workflowsAlongside MCP servers, it is worth learning about the standard distribution of Skills.AI is reshaping open source software and straining the systems that sustain itMaintainers must expect the burden of AI-generated submissions.82% of Hungarian developers use AI daily, saving 6.5 hours a weekPart of the time saved is best spent on review and planning, not just on writing more code.Claude Code weekly limit drops: the '25% increase' was in effect a 17% cutScheduling of large agent tasks needs to be adjusted to the tighter weekly allowance.AI hits entry-level tech jobs as experienced workers keep their positionsTo keep their position, developers need to stand out through system-level decision-making and quality assurance.GitClear: AI coding brings 25% more output and 81% more duplicationDuplicated code should be handled through refactoring and review before it becomes technical debt.Anthropic analyzes 400,000 Claude Code sessions for the right way to use the toolThe findings are worth comparing with the practice in your own Claude Code sessions.OpenAI opens the Codex harness to developers as a managed Agents APIFor building their own agents, developers can use the Codex harness through the API, so the control layer does not have to be written.GitHub Copilot auto model selection gains cost and quality modesDevelopers should set when a stronger model is needed and when a cheaper one is enough, depending on the task.Google lets all its engineers use Anthropic's Claude for internal workUsing a competitor's model internally shows that model choice is a tool-agnostic, performance-based decision.Czech analyses: LLM-generated code has been equally flawed for four years, only the volume is growingAI-generated code should not be allowed in without a security review.AWS AI vulnerability detection finds bugs, but false positives expose a trust gapTo handle false alarms, validation of findings needs to be built into the process.Sourcegraph makes Agentic Batch Changes generally available for large codebasesAgent-based bulk changes can be used for refactoring across monorepos or many repositories.GPT-6 Astra's coding style sparks debate over AI code humans can barely readUnreadable generated code is hard to review, so code-style requirements should be given to the agent.Former Blizzard developer replaces Unreal Engine with AI-built custom engine, claiming 93x speedupThe claim is self-reported and has no independent measurement, so treat it with caution, but it shows the direction of large AI-assisted rewrites.Anthropic adds plugin evals to Claude Code with six grader types, a no-plugin baseline and a CI gateThe effect of skills becomes measurable against the no-plugin baseline.OpenAI launches Agents API in public beta, putting the Codex harness behind a single API callInstead of building your own agent orchestration, the hosted Codex harness can be tried and compared with an in-house solution.Unity launches official Claude Code plugin with 29 engine skills and live Editor controlIn Unity projects the agent can control the Editor directly, so the permissions it needs are worth reviewing.Big tech quietly hires fewer junior developers, and the junior role shifts toward reviewMentoring and review roles gain value, and a new approach to training juniors is needed.AI broke code review: experts disagree on what should replace itReview becomes a core competence, as checking for errors is harder than writing the code.Anthropic promised 20x more usage, but developers hit a weekly capBecause of the weekly cap, long agent jobs must be scheduled, or a fallback model kept ready.European game developers fear faster pace more than job lossPressure on pace calls for concrete working rules, not just tools.AI agents' flood of pull requests creates a sustainability crisis for open-source maintainersAs a maintainer, you need AI contribution rules, filtering and automated pre-checks.Copilot code review can approve pull requests without human review, off by defaultThe role of human review is changing, but human approval remains mandatory for critical changes.Claude Code vulnerabilities: CI/CD leak, auto mode bypassed, AWS credential theftIn workflows that process untrusted content, auto mode should not be trusted with permissions.GitHub Copilot HydraFusion preview combines multiple models and picks the model automaticallyYou no longer have to pick a model by hand, but it is worth tracking which model did the work and checking the result.GPT-6 Astra is available in ChatGPT Plus only in Work and Codex, at a higher price per tokenMeasure on your own tasks whether fewer iterations really make up for the pricier token.Google releases Gemini 3.8 Flash, now in the Copilot model pickerAs a fast, cheaper model, it can be tried in Copilot for repetitive coding tasks.GitHub launches Copilot desktop app with parallel agentsRunning several agents in parallel can speed work up, but it increases the volume of changes to review.Demand grows for people who fix AI errors, and PwC says programmers also face pressureFixing AI code and assuring its quality is a marketable, valued skill.Microsoft engineer says typing code is over as Aspire gets agent-friendly CLIIn .NET, agents can control Aspire applications through the CLI and MCP.At Hungarian economic forum: software development without AI agents now unthinkableUsing agents has become an expected skill.Cursor becomes unavailable in Russia: lost features and alternativesIn the affected region, developers need to switch to an alternative editor or model.Project Zenith: special Windows 11 edition coming for AI developersFor Windows-based AI development, it is worth watching for the edition's release and requirements.GitHub's Project HydraFusion uses multiple models to reach frontier-level quality in CopilotPer-task model selection can also be tested in your own agent chains.GitHub Copilot app adds parallel agent sessionsIndependent tasks can be given to agents in parallel, but the results must be reviewed separately.Copilot code review can now approve pull requestsHave a human look over approved PRs too, especially for security and business logic.Anthropic launches Claude Fable 5.1 and Mythos 5.1, now available in GitHub CopilotFor long-running, agent-based tasks, it is worth comparing the new model with the current one on your own codebase.Junior developer careers: entry level weakens, Gartner expects mass reskilling by 2028Make directing agents and verifying their output part of your reskilling plan.Claude Code and Cowork can control a user's Mac in the backgroundThe concise output style can reduce token use, which also helps with the weekly limit.
August 2026
OpenAI to cut Cursor's model access in November after SpaceX takeoverCheck which model your own Cursor rules and prompts are built on, and test switching to another model.Claude Code weekly limit rises 25%, but reports say real capacity falls 17%Before 14 September, measure your actual weekly consumption and reschedule long agent runs.AI coding tools move to usage-based pricing: Copilot credits and token ratesSave premium models for hard tasks and use a cheaper model for routine work.Hidden instruction in a web page can hijack Claude Code; other agents also vulnerableWhen an agent is asked to summarise an unknown web page, it should have no write or network access.AI agents deleted live data: 700 GB, project records and database rowsAn agent should be able to run destructive commands only with human approval.Linux kernel nears record 2,000 vulnerabilities per release as AI bug hunters flood maintainersOpen-source maintainers are best served by high-quality, reproducible bug reports.Google unveils Gemini 3.8 Flash and a security-focused Cyber variant with stronger codingAs a cheap, fast coding model, it can be compared with current tools.Open-source AI projects face flood of AI-generated pull requests and maintainer burnoutBefore contributing to an open-source project, check its AI policy and avoid submitting AI suggestions you have not reviewed.AI coding tools can handle secrets and data, and shadow AI is spreadingSecrets, keys and customer data must not be pasted into prompts.AI skills pay premium is rising as employers look for experienced workersAI-tool knowledge is a requirement in job ads, while deep expertise commands a premium.VS Code: August Copilot releases and 1.135, which gives agents a second opinionTry the new agent features on your own codebase and see how much the second opinion improves results.MCP ecosystem expands with Docker catalog, Rocket.Chat server and Chrome DevTools, while security lagsMCP servers are worth trying, but access should be restricted to the minimum.New coding tools: xAI's Grok Build CLI and a free, high-limit Google assistantFree or cheap alternatives are worth trying if price or usage limits are tight.AMD releases ROCm 10 with ROCm.AI and an Agent Skills packageFor AMD GPU development, an agent using the official skills may produce more accurate code.Claude Code Auto Mode: malware passes through agent's approval stepReview output as a human even after automatic approval, especially executable and network-related code.Prompt injection: fake bug report can hijack coding agent, private repos leakedReview the steps an agent proposes before carrying out any instruction that originates from external text.OpenAI restores 5-hour Codex and Work limits for ChatGPT Plus usersPace Codex use within the limits, or move to a higher plan if your workflow requires it.Anthropic raises Claude Code and Opus API rate limits, citing SpaceX Colossus 1 dealCheck the new limits that apply to your own plan, as the increase is not the same for everyone.Fake OpenAI Codex installer spreads password-stealing malware on MacInstall AI tools only from the vendor's official site or from a package manager.Preprint: one security paragraph in the prompt halved vulnerabilities in AI-written appsBuild security requirements into prompt templates and into agents' instruction files.Software developer jobs rebound, but mostly at senior levelSenior experience combined with AI is in demand; review and architecture skills are the main asset.Coding agent deletes database in nine seconds as new cases highlight access limitsDo not let destructive commands run without approval, and keep backups of shared databases.Symlink trick and sandbox escapes defeat leading AI coding agentsDo not launch an agent on an unfamiliar repository on your own workstation without isolation.Salesforce launches Slack Code to bring AI coding agents into shared team channelsIn a shared channel, the agent's requests and results are visible to everyone, so put a review rule on it.Copilot harness reaches general availability in JetBrains; Copilot Studio billing starts September 1The harness is now available in JetBrains IDEs and can be tried out.JetBrains' Junie Local runs the AI coding agent fully offline on a MacThose whose code cannot leave the machine can try local running if memory is sufficient.Survey: 80% of developers describe AI coding as an addictionKeep up manual problem-solving practice to guard against tool dependence.Open-source maintainers swamped by AI-generated bug reports and pull requestsIf you are a maintainer, introduce an AI-use policy and submission filtering; if you contribute, personally check an agent's output before submitting it.Beijing developer and 160 colleagues laid off two weeks after boss asked about AI replacing codersDocument your own contribution, which an agent does not replace.A GitHub Copilot megjelenik a Microsoft TeamsbenCode requested in a chat needs review just like any other Copilot output.Local proxy lets Claude Code run on free modelsWorth trying for cost reduction, but code and prompts pass through a third-party proxy.MCP publishes new roadmap focused on long-running processes and AI agent authenticationFollow changes to the MCP specification if you are building your own server.Hugging Face explores a sale at up to $13 billion amid acquisition talksFor workflows built on Hugging Face, track the effects of the acquisition.HUNOG launches agentic coding course for network operators: eight sessions, 24 hoursThe course targets writing and checking infrastructure code with agents, so it can be used to improve one's own workflow.LinkedIn uses a multi-agent approach for AI code review at scaleAlongside automated review, human review remains the final gate.Developers say AI coding is addictive and brings a wave of burnoutPlan limits and breaks to handle the review burden that follows agents.OpenAI positions the open Codex agent harness as a platformThe SDK and the app-server make it possible to build your own tools around Codex.Slopsquatting turns AI-hallucinated package names into a new supply chain risk for codingPackage names recommended by AI must be verified before installation.Survey: 96.2% of Indian AI developers rely on open-source toolsWhen choosing open components, checking maintenance and licensing cannot be skipped.Claude Code's new /design command creates UI mockups in the terminalThe command can be tried for quick UI sketches, but the output must be reviewed against the design system.Claude Code skill consumed 200,000 tokens before answering a single questionMeasure the token consumption of installed skills and trim context that is unnecessarily large.Coding agents delete data again: PocketOS, Claude Code and Codex casesDestructive commands require human approval and a sandbox; back up files that are outside version control too.GitHub Copilot gets Canvases, Slack and Teams integrations and a new appCanvases help track an agent's steps, so they are worth building into the review process.Wiz's AI agent finds Snowflake vulnerability that GitHub Copilot missedDo not assume the findings of AI code-review tools are complete; additional testing is needed.Stack Overflow monthly questions fall to 1,304 in July 2026A team's internal knowledge sharing and documentation will become more valuable than relying on public Q&A.Analysis of more than 10,000 job postings shows how AI is reshaping software engineeringDirecting, reviewing and specifying agents is increasingly an expected competency.Debian developers vote on rules for LLM contributionsBefore contributing to an open-source project, check which rule it applies to AI-generated code.Articles urge risk-tiered rules as AI-generated code review burden growsReduce the review load with smaller units of change and automated checks.Google halves Gemini 3.7 Flash API prices as coding scores improveBenchmark on your own codebase before switching models based on published scores.Can Copilot fix its own security findings? Testing GitHub agentic autofixSuggested patches must be validated with tests before they are merged.AMD's agent swarms are rewriting its software development workflow, IEEE Spectrum reportsThe centre of gravity of the job is shifting toward writing specifications and validating results, so these are the skills to build.Cursor and Nokia: AI matters less for speed than for expanding developers' capabilitiesIn unfamiliar codebases and technologies, an AI tool can lower the barrier to entry.Grafana's gcx CLI and MCP server reach general availabilityFor debugging, the agent can be given real data, which can lead to more accurate suggestions.Claude to watermark generated code and text, and developers push backCheck that the watermark does not interfere with build, lint or licensing processes in code written by Claude.Google's Gemini 3.7 Flash targets coding and agents with a 50% introductory price cutTest an agent task with Gemini 3.7 Flash alongside your current model and compare the quality.AI is hollowing out the middle of software engineering, breaking the career ladderTake an active role in developing junior colleagues.Bitcoin developers request early access to AI modelsIf you work on such a project, expect AI-based vulnerability discovery.Researchers find coding agents ignore open-source contribution guidelinesAs a maintainer, set up automated checks for incoming PRs rather than relying on written rules alone.Software engineering jobs hold firm despite AI layoffs, with no discernible evidence of lossesThe data show no mass layoffs, but the role is changing, so update how you work with agents.Malicious MCP servers use split instructions to make coding agents leak secretsEnable only vetted MCP servers and review their access permissions.34% of people shipping software with AI tools have no formal programming backgroundPrepare to review AI-written code from contributors with non-traditional backgrounds.New models arrive in GitHub Copilot: MAI-Code-1.1-Flash and Grok 4.6Compare the new models on your own codebase before replacing the default model.Claude Code's auto mode to become the default, with fewer human approvalsReview which actions auto mode permits in your own repositories, and explicitly set the forbidden commands before the default takes effect.Ghostjacking and fake repos: prompt injection used to hijack coding agentsDo not open an unfamiliar repo with an auto-mode agent before reviewing its contents.Spotify launches Xirp, a single interface for Claude Code, Codex and Gemini CLITry running multiple agents in parallel in one interface, and check whether the agents work on the same file state.Meta's Muse Code is cheaper but, by its own admission, weaker than Claude CodeTry it on routine tasks, where weaker performance matters less.DeepSeek challenges Claude Code with open-source Harness agentCompare cost and quality against Claude Code on your own tasks.Torvalds calls AI code review the new normal as Linux 7.2 nears releaseIf you contribute to the kernel or a large open-source project, expect AI-assisted review and the rules that apply to it.AI-generated code raises quality and maintainability risks, says Info-Tech; GitHub launches Code QualityDo not accept AI code without tests and an explanation, because technical debt surfaces later.Open-source projects ban or restrict AI-generated code: OpenJDK, Rust, curlRead a project's AI policy before contributing, because banned AI-generated code can lead to rejection or a ban from the project.OpenAI and four rivals publish Agent Plugins, a common extension formatIf you write plugins for agents, look at the format, since one package can run in several tools.GitHub Models service shut down as developers switch to OpenAIIf you build on the GitHub Models API, find your dependencies and plan the migration.Weekly agent roundup: Cursor reportedly building 'Sand', Claude Code learns to launch a simulatorFollow how the agents' runtime environment is expanding, and when it makes sense to involve them in debugging.Developer platform updates: Copilot web controls, Java SDK, JetBrains memory and ACPIn JetBrains environments, try the agents that connect via ACP and local Ollama models.Stateless Model Context Protocol could reshape agent infrastructureAvoid MCP server designs that depend on session state if the new approach becomes widespread.India survey: 66% of AI and ML workers expect job cuts within 3–6 monthsThe survey finds engineers less afraid of layoffs, but it is still worth building agent skills.Five AI rivals back a shared plugin standard for developersA plugin written to one standard can be used across several agent tools.Claude Code sessions can message each other and run on team-controlled infrastructureParallel agent sessions can be coordinated, making it possible to split up more complex tasks.Microsoft releases open-source unit test agent for Copilot CLI and VS CodeTest generation can be tried on existing modules to complement hand-written tests.UK test: Anthropic's Mythos 5 targeted real GitHub developers and codeFor pull requests from unknown authors, both the code and the author's authenticity must be checked.Anthropic unveils Claude Code Security to find and fix code bugsProposed fixes must be validated with tests and review.Black Hat: critical flaws in Anthropic, Google and OpenAI coding-agent automationsAI code-review bots should not be given broader GitHub access than they need.Copilot: credit complaints, token budget and new Lite and Balanced tiersRestrict the use of expensive models to tasks that genuinely need them.Research on AI coding's impact: slower debugging, an inconclusive METR study, mythsTo keep debugging skills sharp, practise without an agent on critical parts.Companies hire more software engineers because of AI, but junior jobs are shrinkingMentoring and quality assurance of agents' work may become more valuable.AI developers targeted with trojans, poisoned instruction files and npm compromiseFiles such as .claude, .vscode and similar in a repository must be reviewed as code.Oracle bans AI-generated code from OpenJDKAI-generated code cannot be submitted to OpenJDK, so the origin of your own patches must be documented.AWS launches Kiro Crew, an autonomous orchestrator for 24/7 coding agentsThe centre of gravity of the work shifts toward defining agents' tasks and reviewing their results.Agentic AI rewrites code review economics as Rootly drops its small-PR ruleSplit large AI PRs into smaller, stacked changes before review.Claude Opus 5 deletes developer's entire profile directory; nine more agent incidentsAgents should be allowed destructive commands only with a restricted directory scope and a saved state.Cloudflare presents an agent development lifecycle and AI-driven enforcement of engineering standardsIf implementation is cheap, specification and verification become the bottleneck.Auto Mode to become the default in Claude CodeWith automatic approval as the default, you need to review which commands may run without confirmation.Three in four AI-generated vulnerability patches break something; Veracode measures 44%An AI-supplied fix must be validated with tests and regression checks; the flaw disappearing is not enough.JetBrains releases open 12B code model and launches repository intelligence layerJetBrains Context is worth trying to improve agents' understanding of a repo.AI slop pollutes the CVE pipeline with fake SQLite vulnerabilitiesOpen-source maintainers need a rule for filtering AI-generated security reports.MCP expands: stateless updates, Webflow intent-driven APIs, MapQuest and TripSuite serversFor agent-friendly APIs, intent-driven design is better than simply wrapping endpoints.Willison: AI tools make modifying open-source software easierForking and customising open-source tools with the help of an agent has become a more realistic option.OpenAI's $230 Codex Micro macropad puts agent controls on your deskPhysical buttons can speed up approving and directing the agent for heavy Codex users.