@@ -59,6 +59,18 @@ Next, we'll discuss the uncontrollable factors we discovered in actual operation
5959#### Environment Instability and Changes
6060
6161- ** Anti-crawling mechanisms and CAPTCHAs** : Google search, shopping websites showing "blocked by the website as robot", verification challenges
62+
63+ <div style =" display : flex ; justify-content : center ; align-items : center ; gap : 20px ; margin : 25px auto ; max-width : 800px ; flex-wrap : wrap ;" >
64+ <figure style =" flex : 1 ; min-width : 300px ; max-width : 380px ; text-align : center ; margin : 0 ;" >
65+ <img src="/blog/osworld-verified/access_denied.png" style="max-width: 100%; height: auto; border: 1px solid #ddd; border-radius: 8px;">
66+ <figcaption style="text-align: center; font-size: 14px; color: #666; margin-top: 8px;">Access Denied - Websites blocking automated agents</figcaption>
67+ </figure >
68+ <figure style =" flex : 1 ; min-width : 300px ; max-width : 380px ; text-align : center ; margin : 0 ;" >
69+ <img src="/blog/osworld-verified/amazon_captcha.png" style="max-width: 100%; height: auto; border: 1px solid #ddd; border-radius: 8px;">
70+ <figcaption style="text-align: center; font-size: 14px; color: #666; margin-top: 8px;">CAPTCHA Challenge - Human verification requirements</figcaption>
71+ </figure >
72+ </div >
73+
6274- ** Network access restrictions** : 403 IP blocking issues (Steam connection timeout, NBA.com geo-restrictions)
6375- ** Dynamic content changes** : Website UI overhauls causing DOM structure changes
6476 - e.g., Apple comparison page URL parameter changes, Budget.com introducing CAPTCHAs from some point
@@ -114,7 +126,7 @@ Many tasks exhibit complex temporal dependencies where proper initialization req
114126 <figcaption style="text-align: center; font-size: 14px; color: #666; margin-top: 8px;">Method 1: Sequential paragraph selection - selecting paragraphs individually</figcaption>
115127 </figure >
116128 <figure style =" flex : 1 ; min-width : 300px ; max-width : 380px ; text-align : center ; margin : 0 ;" >
117- <img src="/blog/osworld-verified/2_linespace_line_by_line_2 .png" style="max-width: 100%; height: auto; border: 1px solid #ddd; border-radius: 8px;">
129+ <img src="/blog/osworld-verified/2linespace_line_by_line_2 .png" style="max-width: 100%; height: auto; border: 1px solid #ddd; border-radius: 8px;">
118130 <figcaption style="text-align: center; font-size: 14px; color: #666; margin-top: 8px;">Method 2: Bulk selection - selecting both paragraphs together</figcaption>
119131 </figure >
120132</div >
@@ -173,17 +185,6 @@ For tasks we identified as genuinely problematic, we primarily modified only the
173185
174186** Problem** : Websites blocking automated access through CAPTCHA, IP restrictions, or bot detection.
175187
176- <div style =" display : flex ; justify-content : center ; align-items : center ; gap : 20px ; margin : 20px auto ; max-width : 800px ; flex-wrap : wrap ;" >
177- <figure style =" flex : 1 ; min-width : 300px ; max-width : 380px ; text-align : center ; margin : 0 ;" >
178- <img src="/blog/osworld-verified/access_denied.png" style="max-width: 100%; height: auto; border: 1px solid #ddd; border-radius: 8px;">
179- <figcaption style="text-align: center; font-size: 14px; color: #666; margin-top: 8px;">Access Denied - Websites blocking automated agents</figcaption>
180- </figure >
181- <figure style =" flex : 1 ; min-width : 300px ; max-width : 380px ; text-align : center ; margin : 0 ;" >
182- <img src="/blog/osworld-verified/amazon_captcha.png" style="max-width: 100%; height: auto; border: 1px solid #ddd; border-radius: 8px;">
183- <figcaption style="text-align: center; font-size: 14px; color: #666; margin-top: 8px;">CAPTCHA Challenge - Human verification requirements</figcaption>
184- </figure >
185- </div >
186-
187188** Solutions Deployed** :
188189- ** Proxy infrastructure** : Added ` proxy ` field support for websites with aggressive anti-crawling
189190- ** Alternative website selection** : For heavily protected sites (e.g., SeatGeek → Ticketek, TripAdvisor proxy issues), switched to functionally equivalent alternatives
0 commit comments