Issue 3
Viewpoint
tool change · global
Cursor launches Origin code hosting platform as GitHub suffers outage
Cursor has unveiled Origin, a code hosting feature in early beta, so the AI editor now offers its own repository hosting. Reports say the GitHub outage points to an opening in the AI coding race.
Who is affected and how
- tech lead
- Try Origin on a small project and measure how much is gained from tighter integration between the editor and the repository.
- CTO
- Tying code hosting and the AI tool to one vendor creates a new supplier dependency that should be weighed in strategy.
incident · global
GitHub outage of more than three hours takes down Copilot and build systems
GitHub went down for more than three hours on Monday morning, taking with it the website, code review tools, automated build systems and Copilot. The provider says the service is now operating…
tech leadIn light of the incident, revisit SLA assumptions and Copilot dependency in sprint planning.CTOThe outage confirms the risk of relying on a single provider for code hosting, so a multi-provider strategy is worth examining.survey · globalJetBrains survey: 90% of professionals use an AI coding agent weekly, Claude Code leads
According to JetBrains, 90% of professional developers used an AI coding agent in some form at least weekly between May and July 2026. In heise's summary Claude Code is the most popular agent, while…
CTOAgent use is widespread, so the realistic goal is managed adoption and policy rather than a ban.tech leadYour team's agent use needs to be assessed and regulated before shadow use develops.junior developerUsing agents is a baseline expectation; the ability to verify their output is the distinguishing skill.incident · globalCoding agents delete data again: PocketOS, Claude Code and Codex cases
According to one analysis, a Cursor agent running Claude Opus 4.6 deleted PocketOS's production database and backups in nine seconds. Other reports say Claude Code twice deleted an agent…
senior developerDestructive commands require human approval and a sandbox; back up files that are outside version control too.startup founderEven a small team should require a separate environment and a restore test before granting agents access.security · globalAI-related DevOps incidents nearly triple year over year, GitProtect Lab finds
GitProtect Lab tracked 84 AI-related DevOps incidents in the first half of 2026, roughly a threefold increase year over year. In the attacks, a poisoned file or public comment can trick an…
tech leadDo not treat input from public comments and external files as trustworthy for agents.Hungary
caseRecord exit in Bulgarian AI software market involves a Hungarian investorstartup founderA large exit in the region also bears on the funding and exit options of Hungarian and Central European AI software teams.caseHUNOG launches agentic coding course for network operators: eight sessions, 24 hourssenior developerThe course targets writing and checking infrastructure code with agents, so it can be used to improve one's own workflow.
More key stories of the week
tool change · global
Slack Code brings AI agents and coding together in shared channels
Slack has launched Slack Code, in which teams can work with AI agents in shared channels; the feature is available in all Slack plans from the day of the announcement. It works with agents available…
tech leadVibe coding in shared channels makes work more transparent, but it requires permission and review rules.tool change · globalGitHub Copilot gets Canvases, Slack and Teams integrations and a new app
GitHub Copilot this week used Canvases to make agent workflows visible and controllable, and gained new experiences in Slack and Microsoft Teams. One review found the Copilot app more useful as a…
tech leadBecause coding tasks can be started from chat, you need rules on who may launch agents from team channels.senior developerCanvases help track an agent's steps, so they are worth building into the review process.survey · globalDevelopers say AI coding is addictive and brings a wave of burnout
Reports say 80% of developers state that AI coding is addictive, and that the tools are triggering a new wave of burnout. The articles stress the contradiction between the productivity promise and…
tech leadTeam workload and the rhythm of AI use must be managed deliberately, not just the speed of output.senior developerPlan limits and breaks to handle the review burden that follows agents.security · United StatesWiz's AI agent finds Snowflake vulnerability that GitHub Copilot missed
Wiz's AI agent, called Red Agent, independently discovered and exploited a vulnerability in Snowflake's internal systems. According to Wiz's blog, GitHub Copilot did not catch the flaw.
senior developerDo not assume the findings of AI code-review tools are complete; additional testing is needed.open source · globalOpenAI positions the open Codex agent harness as a platform
OpenAI's post of 19 August presents the open-source Codex harness as infrastructure for custom applications. Guides discuss the differences between codex exec, the SDK and the app-server, as well as…
senior developerThe SDK and the app-server make it possible to build your own tools around Codex.case · AsiaCyberAgent's ¥30,000 monthly AI allowance boosts Claude Code and Codex use
An AI support programme at Japan's CyberAgent, covering about 1,200 engineers, increased the use of Claude Code and Codex over one year. The article also examines which tools ended up less used.
CTOSubsidising tool costs as a monthly allowance can raise adoption, but plan for usage measurement and a review of the tools that get replaced.tech leadThe team should track which earlier tools fall out of use after agents are introduced.tool change · United StatesOpenAI cuts developer pricing for frontier GPT-5.6 Sol model by more than 20%
According to Reuters, OpenAI cut developer pricing for its frontier GPT-5.6 Sol model by more than 20% on August 21. The price cut affects the costs of teams building on the API.
startup founderRecalculate the AI cost per request and consider whether to switch to a cheaper model or put the savings toward margin.CTOThe price change is a reference point in vendor negotiations and in the cost budget for coding agents.incident · globalNobody knows who built free Ox Alpha coding model or where submitted code goes
Ox Alpha appeared on OpenRouter with a one-million-token context window and no price; according to the listing, prompts and outputs may be logged. The model is gaining popularity among developers…
tech leadThe team should be given a clear list of approved models.startup founderA freely available model is convenient, but the risk of business code leaking is high.case · United StatesCursor reported to be part of SpaceX, positioned as a challenger to OpenAI
According to available reports, Cursor is part of SpaceX, which is using the AI startup to challenge OpenAI. Cursor previously moved from assistive tools toward more autonomous coding.
CTOGiven the change of ownership, reassess Cursor's vendor risk, data-handling terms and pricing stability.startup founderFor workflows built on Cursor, keep a backup tool at hand in case of a strategy change.research · globalMETR follow-up study reverses the 19% slowdown result
In METR's first experiment, 16 experienced developers took 19% longer on 246 tasks with AI in mature projects. According to an analysis of 17 August, the follow-up study flips the sign, so the old…
CTOInterpret productivity claims by task type and codebase maturity rather than relying on a single number.tech leadMeasure the effect on your own team before committing to a roadmap built on AI tools.security · globalSlopsquatting turns AI-hallucinated package names into a new supply chain risk for coding
AI assistants can recommend packages that do not exist, which attackers can then register. According to USENIX Security 2025, 19.7% of the suggested packages did not exist, and 43% recurred across…
senior developerPackage names recommended by AI must be verified before installation.case · globalCursor admits its new coding model is built on Moonshot AI's Kimi
Cursor has admitted that its new coding model is based on Moonshot AI's Kimi model. The disclosure puts the transparency of the origins of models behind tools in the spotlight.
CTOAdd the model provenance of suppliers to the vendor assessment questionnaire.In brief
releaseClaude Code now available with a self-hosted runner for Docker and GitHub ActionssurveyStack Overflow monthly questions fall to 1,304 in July 2026junior developerProblem-solving increasingly relies on AI assistants and documentation, so critically checking output matters.senior developerA team's internal knowledge sharing and documentation will become more valuable than relying on public Q&A.job marketAnalysis of more than 10,000 job postings shows how AI is reshaping software engineeringjunior developerBuild the AI skills named in job postings deliberately and show them in your portfolio.senior developerDirecting, reviewing and specifying agents is increasingly an expected competency.tech leadAlign hiring profiles with the skills needed alongside AI tools.job marketWEF: AI is reshaping the path from junior to senior developerjunior developerDeliberately learn system design and debugging, because agents take over the simple tasks.tech leadSet up a mentoring and knowledge-transfer programme so juniors do not simply accept AI output.securityRansomware operator used Claude Code to strengthen attacks, report saysCTOThe risk of AI misuse should be included in threat modelling and in vendor assessments.caseAsana replaces a legacy test system in two weeks with Codex, at about $12,000QA testerTest-system migration can be accelerated with agents, but coverage and correctness must be confirmed by human review.tech leadLong, mechanical migrations are good candidates for an agent pilot, with cost and deadline estimates.open sourceDebian developers vote on rules for LLM contributionssenior developerBefore contributing to an open-source project, check which rule it applies to AI-generated code.tech leadIt is worth writing an AI-contribution policy for your own open-source projects too.caseCognition in talks on new funding round at $40 billion valuationstartup founderThe capital-market signal shows investor interest in autonomous coding agents remains strong, but competition is also crowded.CTOGiven market consolidation and large capital injections, both price and long-term viability should be weighed when choosing a supplier.caseArticles urge risk-tiered rules as AI-generated code review burden growstech leadIntroduce a risk-tiered review rule so that large AI-generated diffs do not slip through without being reviewed.senior developerReduce the review load with smaller units of change and automated checks.QA testerTests and acceptance criteria will play a more prominent role in verifying AI-generated code.tool changeClaude Code's new /design command creates UI mockups in the terminalsenior developerThe command can be tried for quick UI sketches, but the output must be reviewed against the design system.junior developerTerminal-based mockups speed up the start of frontend tasks, but decisions should rest on the team's existing components.releasexAI's Grok 4.6 model now available in GitHub Copilot for enterprise developerstech leadDecide in the admin settings whether to enable the model, and compare it with the others on smaller tasks.tool changeReplit launches Free Mode powered by OpenAI's GPT-5.6 Luna modelstartup founderIt can be tried at no cost for building prototypes, but usage terms and data handling should be checked before scaling.junior developerA free entry point for quick experimentation, though checking the generated code remains the developer's responsibility.tool changeAnthropic extends Claude Code weekly usage boost through August 31tech leadBefore the bonus expires, assess the team's actual usage so that limits returning to normal do not stall work.startup founderThe higher allowance lasts only until the end of August, so it should not be treated as permanent when planning costs.surveySurvey: 96.2% of Indian AI developers rely on open-source toolsCTOGiven the dominance of open tools, managing licence and supply-chain risk is a basic requirement.senior developerWhen choosing open components, checking maintenance and licensing cannot be skipped.tool changeAnthropic tokens cost more on Vercel, and non-English prompts may add hidden costsstartup founderCheck the actual token price of using a model through a platform, and the cost of non-English prompts.tech leadMeasure token cost by platform and by language before budgeting.caseClaude Code skill consumed 200,000 tokens before answering a single questionsenior developerMeasure the token consumption of installed skills and trim context that is unnecessarily large.tech leadThe team's shared set of skills should go through an approval and cost-tracking process.releaseGoogle halves Gemini 3.7 Flash API prices as coding scores improvestartup founderThe cheaper Flash model could be an alternative to pricier models for simpler coding tasks, so it is worth benchmarking on your own workloads.senior developerBenchmark on your own codebase before switching models based on published scores.job marketIs AI reducing employment for software coders? A look at the coding labor marketjunior developerIn the contest for entry-level positions, verifiable work done with AI tools is what sets candidates apart.CTOIn headcount planning, plan for roles to be reshaped rather than replaced.securityCopilot tricked into revealing how it can be hackedsecurityTexas student exposes rogue AI hacking attempt hidden in GitHub supply-chain trapreleaseAzure DevOps remote MCP server reaches general availability without Claude, ChatGPT or Cursor supporttech leadWhen choosing an MCP server, check which agents are supported.tool changeCan Copilot fix its own security findings? Testing GitHub agentic autofixsenior developerSuggested patches must be validated with tests before they are merged.tool changeWarp launches Warp Factories, bringing the AI-factory model to software developmenttech leadFor workflows handed to agents, define gates and metrics before the team works in a factory model.caseAMD's agent swarms are rewriting its software development workflow, IEEE Spectrum reportssenior developerThe centre of gravity of the job is shifting toward writing specifications and validating results, so these are the skills to build.tech leadTasks need to be broken down so that agents can work in parallel, with human review acting as a gate.caseCursor and Nokia: AI matters less for speed than for expanding developers' capabilitiesCTOWhen justifying a rollout, treat the new roles and capabilities it covers as a measurable goal alongside speed.senior developerIn unfamiliar codebases and technologies, an AI tool can lower the barrier to entry.tool changeGitHub Copilot's monthly usage hits record high after new pricing modelCTOUnder the new pricing, track teams' actual consumption, because costs can rise with usage.tech leadPermissions and usage need to be monitored to avoid unexpected bills.caseLinkedIn uses a multi-agent approach for AI code review at scaletech leadThe multi-agent review architecture can serve as a model, but it should first be tested on a small area of the codebase.senior developerAlongside automated review, human review remains the final gate.caseJapanese insurer gave 3,000 staff GitHub Copilot, but only basic features were usedCTOHanding out licences is not enough: without training, usage measurement and targeted workflows, the cost does not pay off.tech leadWork out usage patterns for advanced features within the team, and measure adoption.caseVibe-coding startup Anything reaches $2 million in annualised revenue within two weeksstartup founderThe market for tools aimed at non-developers is growing fast, but competition and quality risk are also significant.junior developerVibe-coding tools widen the population that writes code, so quality-assurance skills gain value.tool changeDeepSeek quadruples its rates and takes on Claude Code with a new modelstartup founderCost models built on DeepSeek need to be recalculated with the higher rates.CTOThe price advantage of the cheap alternative may shrink, so plans to switch models should be re-evaluated.securityZhipu says its GLM-5.3 coding model developed cyber-offence skills faster than expectedreleaseGrafana's gcx CLI and MCP server reach general availabilitysenior developerFor debugging, the agent can be given real data, which can lead to more accurate suggestions.caseChinese AI code-review startup becomes a unicorn at a 10.8 billion yuan valuationstartup founderCode review is a market that attracts capital, but the field is crowded with competitors.CTOThe choice of AI review tools is widening, so it is worth starting vendor evaluation now.caseNvidia pays $6 billion to license Poolside's models and invests $1 billion in the startupstartup founderThe close tie between a hardware maker and a model developer signals consolidation in the coding-AI market.CTOThe ownership background of a model supplier can be a vendor-risk consideration.open sourceHugging Face explores a sale at up to $13 billion amid acquisition talksCTOA change of ownership at the open model hub could affect the availability and terms of models, so plan a fallback source.senior developerFor workflows built on Hugging Face, track the effects of the acquisition.