Claude Opus 4.7 crash fixes failed maintainers' tests a third of the time in Northwestern study
Northwestern tests of AI agents on 112 real bugs found a third of Claude Opus 4.7's crash-stopping patches failed maintainers' tests. Many patching benchmarks grade only that crash, so their scores overstate how many bugs are fixed.
Reality
- Evidence58
- Adoption
- Not scored
- Hype gap+5
- Incentives
- Not scored
- Confidence55
buildOne report1 publisher WordPress patches a no-login path-traversal bug that default server settings turn into RCE
WordPress shipped 7.1.2 on 22 September 2026 to patch CVE-2026-87902, an unauthenticated path-traversal flaw rated 9.2 that reaches every release since 4.7.0. CISA listed it as known-exploited three days later.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+20
- Incentives
- Not scored
- Confidence50
buildOne report1 publisher Every wiki that ran MediaWiki External Data before 3.7 needs a hunt for planted PHP shells
Automated exploit attempts hit MediaWiki's External Data extension within a day of the 25 September disclosure of CVE-2026-100382, a CVSS 10.0 flaw. Upgrading to 3.7 closes the entry point but leaves behind any PHP shell an attacker already wrote to disk.
Reality
- Evidence50
- Adoption30
- Hype gap+10
- Incentives
- Not scored
- Confidence45
buildOne report1 publisher Exploit attempts for WordPress CVE-2026-87902 began the day 7.1.2 shipped
WordPress released 7.1.2 on 22 September 2026 to fix remote file inclusion flaw CVE-2026-87902, and the first exploit attempt was recorded the same day. The fix was back-ported to every maintained branch down to 4.7, so an exposed site has hours before scanners find it.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+20
- Incentives
- Not scored
- Confidence40
buildOne report1 publisher Laravel 13.34 lets queued jobs count OOM kills against their exception limit
Laravel 13.33 and 13.34 add opt-in guards for two of the three ways a queued job dies in Docker, worker timeouts and crashes. Neither touches Docker's stop grace period, so a slow model call still gets ten seconds by default after a deploy's SIGTERM.
Reality
- Evidence55
- Adoption
- Not scored
- Hype gap0
- Incentives15
- Confidence55
buildOne report1 publisher Symfony 8 swaps generated DI proxies for PHP 8.4's native lazy objects
Symfony 8 creates its container services as PHP 8.4 native lazy objects, cutting boot time 15-25% for apps with 200+ services, according to a dev.to post. The gain comes from retiring generated proxy classes, so it depends on how much boot time those classes cost you.
Reality
- Evidence30
- Adoption
- Not scored
- Hype gap+35
- Incentives
- Not scored
- Confidence35
Patchstack says attackers are chaining CVE-2026-61979 and CVE-2026-15981 to mint WordPress admin sessions. Only the free edition got an advisory; Standard needs 17.0.6.
Perspective Coverage
5 publishers
- Builder
- Builder 32%
- Operator
- Operator 56%
- Investor
- Investor 12%
Reality
- Evidence70
- Adoption30
- Hype gap+15
- Incentives
- Not scored
- Confidence68
buildOne report1 publisher Haiku labels three of eight patched twins as reachable vulnerabilities
A Kaggle-challenge benchmark called ART scores models on whether they still flag a function after the fix is applied. On eight synthetic pairs, the difference between price tiers showed up only on the patched half.
Reality
- Evidence47
- Adoption12
- Hype gap−5
- Incentives58
- Confidence44
buildOne report1 publisher Adobe's extra year on Magento 2.4.6 runs mostly on an unsupported PHP
Magento Open Source stopped patching 2.4.6 on 11 August 2026, and about seven of the paid edition's twelve extra months fall after PHP 8.2 stops getting security fixes. The licence buys planning time.
Reality
- Evidence64
- Adoption44
- Hype gap+6
- Incentives70
- Confidence58
buildOne report1 publisher Click2Shell turns a 5.3-rated WordPress selector injection into PHP on the server
The WordPress Core bug alone scores 5.3. In a research proof of concept it installs an official catalog theme inside the administrator's browser, the Customizer preview loads that theme's functions.php, and a second bug runs the attacker's PHP.
Reality
- Evidence52
- Adoption25
- Hype gap+15
- Incentives55
- Confidence50
buildOne report1 publisher Laravel's bootstrap cost 316KB of the 0.66MB a warm hello-world request used
A dev.to benchmark on Laravel 13.31.0 and PHP 8.4.22 splits a request into framework bootstrap and loaded data. The much-quoted 20MB to 30MB per-request figure turns out to measure PHP with OPcache disabled.
Reality
- Evidence58
- Adoption12
- Hype gap+15
- Incentives22
- Confidence46
buildOne report1 publisher Bubblewrap hands the PDF binary one directory and no network by default
Escaping an uploaded filename keeps it from turning into a shell command, and it still leaves the parser reading everything the worker's account can read. A dev.to write-up draws the other boundary with Linux namespaces.
Reality
- Evidence40
- Adoption
- Not scored
- Hype gap−10
- Incentives55
- Confidence45
buildOne report1 publisher Laravel 13's ten-minute upgrade hinges on what your tests still call VerifyCsrfToken
Laravel calls the breaking changes in 13 minimal, and the upgrade guide still flags about nineteen of them. Two land in defaults that every application touches: the request forgery middleware and the cache config.
Reality
- Evidence48
- Adoption
- Not scored
- Hype gap+10
- Incentives55
- Confidence45
buildOne report1 publisher Renaming ServiceProvider to ServiceLauncher costs five edits per class in Doppar 4
Doppar 4 drops the Laravel-shaped names that 3.x applications were written against, and it will only run on PHP 8.5. The release post tells 3.x users to move the runtime before touching application code.
Reality
- Evidence48
- Adoption10
- Hype gap+15
- Incentives80
- Confidence55
buildOne report1 publisher One shared hosting login puts ten client sites in a single failure domain
A hosting guide from EniyiSunucum sets one account per client as the minimum boundary for an agency portfolio. The part of it worth keeping is the queueing account of why a neighbour's traffic spike hurts.
Reality
- Evidence30
- Adoption
- Not scored
- Hype gap+8
- Incentives60
- Confidence55
buildOne report1 publisher Rerunning the generator in --check mode makes a stale snapshot fail the build
Anton Brilliantov keeps one .proto per interface in a shared contract repository and generates the HTTP specification from it. The same discipline already governs his environment variable catalog and his metrics snapshot.
Reality
- Evidence34
- Adoption12
- Hype gap+8
- Incentives20
- Confidence45
buildOne report1 publisher Passing $header['alg'] to the verifier lets the token choose which algorithm checks it
A JWT signature only proves integrity when the verifier and the application already agree on algorithm, key and validity window, and in the vulnerable pattern all three are settled by input the attacker sent.
Reality
- Evidence52
- Adoption
- Not scored
- Hype gap+12
- Incentives18
- Confidence56
buildOne report1 publisher Socket finds PolinRider in four dev-* versions of a 700,000-download Packagist package
Socket reports no malicious stable release. The exposure sits with teams that resolve dev-* constraints directly, and with anyone who clones the repository and opens it in a VS Code-compatible IDE.
Reality
- Evidence58
- Adoption25
- Hype gap+12
- Incentives75
- Confidence45
buildOne report1 publisher Previewing an unapproved comment triggers do_blocks() on the entire event page
Wordfence's Argus team found two CVSS-9.8 chains in The Events Calendar. An anonymous comment plus a moderation-hash preview URL reaches OS command execution, and version 6.17.4 closes only one of them.
Reality
- Evidence68
- Adoption
- Not scored
- Hype gap+5
- Incentives55
- Confidence62
buildOne report1 publisher A per-API config block absorbs the pagination differences behind one paginate() call
The author of laravel-rest pointed one Eloquent-style client at 62 public APIs across 470 unmocked scenarios and published the wire logs for three of them. Where the rows and the page total live is declared per API.
Reality
- Evidence68
- Adoption
- Not scored
- Hype gap+16
- Incentives74
- Confidence58
Earlier coverage
- Writing the task text before the research encodes a guess
Build · September 11, 2026 · One report1 publisher
- GOautodial runs an agent's logout parameter through /bin/sh
Build · September 2, 2026 · One report1 publisher
- Once the question needs a cube, you own the parser
Build · August 20, 2026 · One report1 publisher