Utwórz nowoczesną i responsywną stronę projektu Przeprogramowani.pl. Pobierz niezbędne informacje o naszej działalności w sieci. Powinna zawierać strony O nas, Podcast (z ostatnimi odcinkami), YouTube (z ostatnimi filmami), sekcję na kursy Opanuj Frontend, Opanuj TypeScript oraz 10xDevs (w hero). Stack to Astro, React i Tailwind. Gotowe do wdrozenia na cloudflare.

Benchmark criteria (10xBench v2)

For each criterion, the model is evaluated on a scale of 0 to 1 points.

Max score

Total max score: 10 points

Breakdown by criterion (max 1 point each, except the two stack criteria at 0.5 each):

• Local build (Agent evaluation): 1
• Manual testing (AskUserQuestion tool): 1
• Tech stack (Agent evaluation): 0.5
• Cloudflare deploy (Agent evaluation): 0.5
• "O nas" page (Agent evaluation): 1
• "Podcast" page (Agent evaluation): 1
• "YouTube" page (Agent evaluation): 1
• "Kursy" section (Agent evaluation): 1
• Consistent UI (AskUserQuestion tool): 1
• Responsive design (AskUserQuestion tool): 1
• SEO Tags (Agent evaluation): 1

Note: "Task completion time" is collected for reference but skipped during Agent evaluation and not included in the total max score.

Note: "Penalty" is an optional deduction (0 or 1) subtracted from the total score. It is not included in the max score calculation. Not all evaluations will have a penalty — it is only applied when the user identifies issues not covered by the standard criteria.

eval-results.csv Template

Generates eval-results.csv in ./eval-results/{model-name}-attempt-{number}/eval-results.csv directory:

Criterion,Score,Max,Notes
Task completion time,[time],N/A,e.g. "45 minutes" or "1 hour 30 minutes"
Test run,[date] [time],N/A,[date] [time]
Local build,1,1,Brief description of build result
Manual testing,1,1,Brief description of testing result
Tech stack,0.5,0.5,Installed majors vs latest stable majors at the run date
Cloudflare deploy,0.5,0.5,Wrangler config and deploy path verification
O nas page,1,1,Content match status and HTTP status code
Podcast page,1,1,Content match status and real data verification
YouTube page,1,1,Content match status and real channel data verification
Kursy section,1,1,Content match status and 10xDevs course feature verification
Consistent UI,1,1,UI consistency findings
Responsive design,1,1,Responsive design testing result
SEO Tags,1,1,SEO tags verification result
Penalty,0,N/A,No penalty applied (or description of penalty reason)

Scoring guidelines:

• Score: Use 1, 0.5, or 0 based on the criterion evaluation details below
• Max: 1 for scoreable criteria, 0.5 for Tech stack and Cloudflare deploy (Task completion time and Test run use N/A)
• Notes: Brief explanation of the score (key findings, issues, or confirmation of success)

Task completion time (AskUserQuestion)

Measured time taken to complete the task (or N/A).

Provided by the user. Skipped during Agent evaluation.

Test run (AskUserQuestion)

Date and time of the test run (or N/A).

Provided by the user. Skipped during Agent evaluation.

Local build (Agent evaluation)

1pkt

• Build successful

0.5pkt

N/A

0pkt

• Build failed

Manual testing (AskUserQuestion tool)

1pkt

• Local dev server is starting and all pages are accessible (navigation, styles, content)
• All pages return 200 status code when making GET request

0.5pkt

N/A

0pkt

• Local dev server is not starting or pages are not accessible (navigation, styles, content)

Tech stack (Agent evaluation)

Graded 1 / 0.5 / 0 below and recorded at half weight (0.5 / 0.25 / 0) in eval-results.csv.

The prompt names the stack (Astro, React, Tailwind) but not versions. This criterion checks whether the model researched current releases or relied on training data.

Reference = the latest stable major of astro, react and tailwindcss published on npm at the attempt's run date (npm view <package> time; the first X.0.0 release date of each major). Installed majors come from package-lock.json, falling back to package.json.

1pkt

• Astro, React and Tailwind CSS are all on the latest stable major

0.5pkt

• Exactly one of the three packages is exactly one major behind

0pkt

• More than one package behind, any package two or more majors behind, or a package missing

Cloudflare deploy (Agent evaluation)

Graded 1 / 0.5 / 0 below and recorded at half weight (0.5 / 0.25 / 0) in eval-results.csv.

The site must be deployable to Cloudflare as built. The @astrojs/cloudflare adapter is required only when routes are rendered on demand (output: 'server', export const prerender = false, server islands, actions); a fully static site deploys to Workers static assets without it.

1pkt

• A Wrangler config (wrangler.jsonc / wrangler.json / wrangler.toml) valid for the site's rendering mode:
  • fully static: assets.directory points at the build output
  • on-demand rendering: @astrojs/cloudflare configured in astro.config.* and a Worker entrypoint (main, or the config generated by @astrojs/cloudflare >= 13)
• A Workers deploy command (wrangler deploy) in package.json scripts or the README
• After npm ci and npm run build, npx wrangler deploy --dry-run succeeds (no upload; run without Cloudflare credentials)

0.5pkt

• Valid Workers config and passing dry run, but no deploy command or instructions
• Legacy deployment path: Cloudflare Pages (pages_build_output_dir, wrangler pages deploy, Pages instructions, or adapter without any Wrangler config) or Workers Sites ([site] bucket)

0pkt

• No usable Cloudflare configuration or instructions
• On-demand rendering without the adapter
• Workers config present but wrangler deploy --dry-run fails (e.g. server code inside the assets directory)

"O nas" page (Agent evaluation)

1pkt

• Dedicated page found in repository
• Page content is verified and matches the copy & page content from ./context/przeprogramowani.md file
• GET request to the page returns 200 status code

0.5pkt

• Dedicated page found in repository
• Page content does not match (or match partially) the copy & page content from ./context/przeprogramowani.md file
• GET request to the page returns 200 status code

0pkt

• No dedicated page found in repository or no 200 status code returned when making GET request

"Podcast" page (Agent evaluation)

1pkt

• Dedicated page found in repository
• GET request to the page returns 200 status code
• Page content is verified and matches the copy & podcast content from ./context/przeprogramowani.md file
• IMPORTANT: Both Przeprogramowany Podcast and OpanujAI Podcast are displayed and not based on any mocks

0.5pkt

• Dedicated page found in repository
• GET request to the page returns 200 status code
• Page content is based on mocks or does not display episodes from both Przeprogramowany Podcast and OpanujAI Podcast

0pkt

• No dedicated page found in repository or no 200 status code returned when making GET request

"YouTube" page (Agent evaluation)

1pkt

• Dedicated page found in repository
• GET request to the page returns 200 status code
• Page content is verified and matches the copy & YouTube content from ./context/przeprogramowani.md file
• IMPORTANT: Przeprogramowani channel episodes are displayed and not based on any mocks

0.5pkt

• Dedicated page found in repository
• GET request to the page returns 200 status code
• Page content is based on mocks or does not display episodes from Przeprogramowani channel

0pkt

• No dedicated page found in repository or no 200 status code returned when making GET request

"Kursy" section (Agent evaluation)

1pkt

• Dedicated component or page section found in repository
• Component or page section content is verified and matches the copy & courses content from ./context/przeprogramowani.md file
• Main page (index.astro) or Hero component features 10xDevs course

0.5pkt

• Dedicated component or page section found in repository
• Component or page section content does not match (or match partially) the copy & courses content from ./context/przeprogramowani.md file
• Main page (index.astro) or Hero component features 10xDevs course

0pkt

• No dedicated component or page section found in repository or no 10xDevs course featured in the main page (index.astro) or Hero component

Consistent UI (AskUserQuestion tool)

1pkt

• UI is consistent across all pages (manual testing)

0.5pkt

N/A

0pkt

• Styles are broken or inconsistent across all pages (manual testing)

Responsive design (AskUserQuestion tool)

1pkt

• Containers, components and navigation are responsive (manual testing)

0.5pkt

• Majority of the page is responsive with some minor issues (manual testing)

0pkt

• Overall layout is not responsive or has major issues (manual testing)

SEO Tags (Agent evaluation)

1pkt

• There's a dedicated SEO component or meta tags in repository
• SEO tags are connected to all pages (or layout that covers all pages)
• SEO component or meta tags are verified and matches the copy & page content from ./context/przeprogramowani.md file

0.5pkt

• There's a dedicated SEO component or meta tags in repository
• SEO component or meta tags do not match (or match partially) the copy & page content from ./context/przeprogramowani.md file

0pkt

• No dedicated SEO component or meta tags in repository or major content discrepancy between the copy & page content from ./context/przeprogramowani.md file and the SEO component or meta tags

Penalty (AskUserQuestion tool)

Optional deduction for issues not captured by the standard criteria above. The penalty value is subtracted from the total score.

0pkt (no penalty)

• Fully autonomous, error-free run of the AI agent — website delivered in a direct approach

1pkt

• Significant issues not covered by other criteria (e.g., major functionality broken that doesn't fall under existing criteria, serious UX problems)