Codacy @codacy

Most AI code review tools can't safely block a merge today. Many teams are either using a strict but probabilistic bot 🙉 or an advisory one that enforces nothing 🙈. Others pair AI review with deterministic checks that enforce the same result every time. 🦍 We compare 14 tools and show which tools can be trusted to uphold AI code governance in 2026. 🍌 blog.codacy.com/ai-code-review…

Apart from no Codex, the top 3 AI coding tools used are probably not surprising. What might be surprising is that 42% of orgs that use AI coding are running 2 or more coding tools. It turns out, in late 2026, there's still very much an appetite for experimenting different coding tools.

keyboard_arrow_left Previous keyboard_arrow_right Next

This CX software company closed 1,200 security issues, at 90% accuracy, after moving to Codacy and onboarding 1400 developers. Read full article: eu1.hubs.ly/H0xdsjs0

Chief of Engineering at @EvolvTechnology on our AI Inventory. Learn more about it here eu1.hubs.ly/H0xdrC20

We talked to hundreds of software teams about how they actually use AI coding agents. What we found was surprising: most teams are simultaneously running autocomplete in one repo, letting agents build whole features in another, and running loops against their own infrastructure. One other thing is unanimous: everyone is handing agents more of the mechanical work so people can move up to the judgment calls. We mapped this into an AI Coding Maturity Scale. Where does you fit? Read full piece here eu1.hubs.ly/H0xdtHz0

Archiving a repository on GitHub now automatically removes it from Codacy. No more read-only repos padding your metrics or queued for analysis they don't need. Small but real operational win for leaders managing large repo estates. Learn more: eu1.hubs.ly/H0xdslk0

The next leap in code review will come from knowledge bases. The more an agent knows about a specific team's decisions, patterns, and constraints, the better its work. Turning that institutional knowledge into something agents can act on is the cog building the future of code review. Read full piece by Codacy CEO @jaimefjorge eu1.hubs.ly/H0xdmZ30

While many teams bolt AI onto code review, Erik Jost's team at @blackbox_ns built a PR gate that's self-hardening: Claude, Codex, Codacy, and GitHub Copilot review each PR in parallel. Then every repeatable fix becomes a new rule for future runs, unlocking a flywheel of speed and trust in the loop.

Engineers moved from building to orchestrating agents, and human attention shifted to review. But review hasn't scaled the way code generation did. That's where automated quality and security enforcement is needed to keep teams shipping fast but safely. Here's how you can build that system: eu1.hubs.ly/H0wPkk-0

Auditors are starting to ask questions engineering leaders can't yet answer, like which of your shipped code was AI-generated, and who verified it? Here's what to have ready before the audit: eu1.hubs.ly/H0wHZpw0

@jaimefjorge @lucaronin where would you position @tolariamd? :)

Steal this AI maturity scale to guide your team toward loop engineering. 👇 (sound on 🔊) Most teams think they're at one "stage" of AI adoption. They're not. They're at all of them at once. We spent the last few quarters talking to hundreds of software teams about how they really use AI and that surprised us. Most people picture AI adoption as a ladder: autocomplete → prompting agents → "loop engineering," where agents run scheduled tasks, wired into your systems, operating on their own. But the teams we talk to are doing all three at once. Autocomplete in one repo. Agents shipping features in another. Already running loops against their own infra. What's consistent is the direction: more autonomy for agents, hand off the recurring and mechanical work, let people move up to the part that still needs a human. Most teams are further down that road than they realize. More autonomy brings more entropy and runaway cost if you scale agents uncontrollably (we've watched companies do exactly that). Plus real change, like reworking your Git flows to handle looping agents. I pulled the best insights from those conversations into the video. Watch it and you'll probably spot where your team sits.

Verity is now live in public beta. A local, adversarial review layer built for coding agents. It pairs deterministic checks with an independent model, to catch and repair security, quality, and intent gaps after every agent run. Every decision compounds into a markdown knowledge base, so each session starts smarter than the last. Plus, you get live cost visibility across all your agents. It’s free while in beta. Give it a try today and send us your feedback: npm install -g @codacy/verity-cli && verity init Learn more: verity.md

Jaime Jorge @jaimefjorge

3 months ago

Today we’re launching Verity.md in public beta. If you’re working with coding agents, we built this for you. As AI writes more code, developers and engeneering teams lose visibility into quality, context, and cost. Verity adds: → Gates (AI + deterministic reviews) → Memory

2 1 9 869 0

"Are you getting it?" Spoiler: Tomorrow you will.

Maybe someone can help me: what is effectively the difference between using claude code's ultracode (which you can define a goal and it uses workflows with many agents etc) and loop engineering? isn't it effectively the same?

Across our customers' repos, we found several instances where agent instruction files (like Claude.md) contain a hardcoded secret. Be sure to your treat your agent instructions like you’d treat your code. Learn more about how Codacy scans your agent files blog.codacy.com/introducing-ag… s/o @simonkim_nft

PR queues are growing faster than review capacity, but rubber-stamp approvals to clear the backlog isn't the answer. This article covers how to keep PRs moving without lowering your standards. Read full piece eu1.hubs.ly/H0w9_mk0

Fable 5's own internal self-verification is strictly probabilistic (it writes its own tests, reflects on reasoning, and uses vision). Because Fable 5 is essentially “grading its own homework,” it relies on the same neural weights and has the same blind spots that produced the code in the first place. If this sounds less than ideal, it's because it is. Check out our recent article on why coding agents need independent quality gates: blog.codacy.com/why-coding-age…

As our customers' agents got busier, so did our SREs. Over the last few months our infrastructure team restructured the database behind code analysis, cut worker memory consumption in half, and tightened how we allocate compute. The result is the average duration of a code analysis is stable even as the analyses keep climbing. Come at us 🤖

This is for the engineers who have to defend paying down tech debt against feature work. We put together a guide on what's actually worth tracking, how to measure it without it turning into an opinion battle in sprint planning, and how to keep the practice alive past the first month. Read full article eu1.hubs.ly/H0vZ3SF0