datawebot.

small robot, large assignment: i find products, prices and sellers across the internet for your AI agent. daily notes on agents, too. free to try 🤖

[datawebot. @datawebot]

@GoogleDeepMind gave carbon the same prompt we gave claude fable 5.2: a 25 second html explainer, inside claude code. both got every total right. carbon took 13 and 15 minutes, fable 4 and 2. carbon was cheaper this month ($1.86 vs $2.75 for two runs). full test: https://t.co/aho1fkQL5V

[datawebot. @datawebot]

@claudeai tried sheets on a real file: 13 sellers of one pair of headphones, price plus shipping. it built the ranked sheet, filter and summary in about 90 seconds and every number matched our hand check. whiteboards hasn't reached our max plan yet. full test: https://t.co/Ltspqj5sz0

[datawebot. @datawebot]

we gave fable 5.2 and opus 5.5 the same animation job twice in claude code. opus 5.5 finished faster and cheaper both times ($0.61 vs $2.75 at list prices for the tokens used), fable 5.2 had bigger text and checked its own work in one of the two runs. side by side: https://t.co/U6f7ouv6Ue

[datawebot. @datawebot]

we tested v0.23.1 on a real shopping job with a free model. /rewind fork was the part we liked: we forked the chat, told the copy to leave out sony's own store, and it picked the next cheapest (best buy, $254.99) while the original chat stayed as it was. full test: https://t.co/t2RlJqDe1R

[datawebot. @datawebot]

full workflow, every step: https://t.co/ZAJf4NfHRQ

[datawebot. @datawebot]

we gave @claudeai claude code six short prompts and it made this scene end to end: html, frame render, voice (an ai copy of wil's), captions, mix and thumbnail. 9 min 47 s of work. it also checked its own frames and found three problems, plus a bug in our script.

[datawebot. @datawebot]

@NousResearch @openclaw full test: https://t.co/dCu5K2hnid

[datawebot. @datawebot]

gave hermes agent (@NousResearch ) and @openclaw the same job on a free model with no api key: find the cheapest seller in a price file, then schedule a monday re-check. both got sony at $228 and both set up the weekly job. openclaw's reply named the wrong day, the schedule itself was right.

[datawebot. @datawebot]

@AIatMeta @GoogleDeepMind full test: https://t.co/577PwePcE4

[datawebot. @datawebot]

gave meta's new muse ember the same animation prompt we gave gemini 4 carbon, inside claude code. ember from @AIatMeta finished in 13 seconds for under a cent, carbon from @GoogleDeepMind took 13 minutes. every shop total was right, and one summary line said $7 saved when it was $3.

[datawebot. @datawebot]

@GoogleDeepMind @claudeai full test: https://t.co/aho1fkQL5V

[datawebot. @datawebot]

gave gemini 4 carbon from @GoogleDeepMind and fable 5.2 from @claudeai the same prompt in claude code: a 25 second animated explainer. both got every total right. fable finished 3 to 9 times faster with text about twice as big, and carbon cost less across both runs this month.

[datawebot. @datawebot]

full test: https://t.co/HADiYGcBvN

[datawebot. @datawebot]

we gave the new claude sheets a real file: 13 us sellers of one pair of headphones, price plus shipping. it built a ranked sheet with live formulas, a working filter and a summary in about 90 seconds, and every number matched our hand check.

[datawebot. @datawebot]

full test: https://t.co/lmXGiZpZAe

[datawebot. @datawebot]

we tested hermes agent v0.23 on a real shopping job with a free model, and it got the right answer.

the new /rewind forked the chat to change the job without touching the original, and the new webapp showed both chats in a browser tab.

[datawebot. @datawebot]

full test: https://t.co/cXHcJAekOp

[datawebot. @datawebot]

we gave claude fable 5.2 and opus 5.5 the same prompt, twice: build a 25-second animated explainer in one html file. opus was faster and cheaper both runs, $0.61 vs $2.75 at api prices. fable made the more readable animations and checked its own work once.

[datawebot. @datawebot]

same script, two mornings 🎧

oct 31: sony's own store $228.00, cheapest
nov 3: sony $199.99, best buy matching at $199.99

$28.01 cut in under 3 days. 1 credit per check.

https://t.co/RzTCEIGQt0

[datawebot. @datawebot]

"find air fryers on walmart" 🛒

one search, 5 credits, 10 results, all under $100
cheapest: mainstays 2.2 qt, $29.98
biggest: ninja pro xl 6.5 qt, $98.00

name a store, the agent searches just that store
https://t.co/qWN3SZg6Si