18 February 2025

This AI Tool Is Like OpenAI Operator But FREE (Proxy AI is INSANE!)

1Frameworks

7Insights

Frameworks in this episode

Productivity6 steps

Prompt-to-Agent Research Workflow

Turn a conversational brief into a structured browser-agent task

Read →

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Hot Take· 1

Hot Take02:30

Small AI Labs Are Compressing the Price of Intelligence

Convergence reportedly raised only $12 million yet competed with Operator on browser tasks, echoing the pressure DeepSeek placed on larger model providers. The hosts see this as evidence that reasoning and agent capabilities are becoming dramatically cheaper.

Small teams can challenge heavily funded incumbents
Open models and cheaper agents pressure premium pricing
Reasoning capability is rapidly commoditizing
Funding size does not guarantee benchmark leadership

“OpenAI have raised 18 billion, and this company has only raised 12 million.”
Kieran · 02:30

“the price of intelligence, the price of reasoning, is coming down at an incredible rate.”
Kieran · 04:00

#ai economics#openai#deepseek#convergence

Explainer· 2

Explainer01:30

Proxy Makes Browser Agents Accessible Without a $200 Plan

Proxy is presented as a free browser agent with a $20 monthly pro option, while OpenAI Operator was limited to OpenAI's $200 tier. The hosts argue that this price difference materially lowers the barrier to experimenting with web-browsing agents.

Proxy offers a free entry point
Proxy has a $20 monthly pro option
Operator was tied to a $200 OpenAI plan
Lower prices can accelerate AI adoption

“And then I saw that proxy was free, and I was like, Oh, I was excited about that, because operators in the $200 a month…”
Kipp · 01:30

“at the moment, operator is available in the $200 tier of open AI. $200 a month proxy is free with a $20 a month pro…”
Kieran · 03:30

#proxy#operator#pricing#ai agents

Explainer02:30

Why Proxy's WebVoyager Result Attracted Attention

The hosts point to WebVoyager, a benchmark spanning 600 web-based tasks, as the reason Proxy's launch generated unusual interest. Convergence's small team reportedly outperformed Operator on this test, suggesting strong browser-navigation capability despite far less funding.

WebVoyager measures agents across 600 web tasks
Proxy reportedly outperformed Operator
The benchmark focuses on web-agent capability
The result amplified interest in Convergence

“there's this web Voyager benchmark, which is, yeah, the agents capabilities across 600 web based tasks.”
Kieran · 02:30

“The reason people are like, holy smokes, is convergence, which is this, small team have outperformed operator.”
Kieran · 03:00

#webvoyager#benchmarks#proxy#browser agents

Story· 1

Story13:30

A Browser Agent Can Assemble a Personalized AI Newsletter

The hosts demonstrate using an agent to review multiple technology and AI sites, rank articles for relevance and engagement, and summarize the strongest items. That research can then be passed to another AI assistant, combined with a format and writing style, and turned into a personalized newsletter.

The agent reviews multiple news sources
Articles are evaluated for relevance and engagement
The output contains summaries of selected stories
Another AI can format the research as a newsletter
Scheduled AI tasks can deliver personalized news each morning

“review all these websites that are AI focused or tech focused, compile a summary of trend and articles and AI and evaluate each article based…”
Kieran · 14:00

“I could pipe it into one of the other AI assistants, apply a format via the prompt and a write in style and create my…”
Kieran · 14:30

#newsletters#personalization#content curation#ai news

Tool· 1

Tool13:00

Using Proxy to Build an AI Influencer List

One host used Proxy to gather a list of 50 prominent AI influencers, including their names, primary platforms, descriptions, and social handles. The result found relevant people but did not reliably provide URLs, illustrating both the promise and current limitations of list-building agents.

The task requested 50 AI influencers
Requested fields included platform and social handle
Proxy identified relevant names
URL completeness was weak
The output could improve with prompt refinement

“I gather a list of 50 prominent AI influencers. I asked for their name, primary platform.”
Kieran · 13:00

“It didn't actually give me their URLs very well, but give me some of the top influencers that have picked that in AI.”
Kieran · 13:30

#influencers#list building#proxy#research

Takeaway· 2

Takeaway10:30

Proxy Browses Faster, but Operator Produces Better Analysis

In the side-by-side account-research test, Operator was estimated to be 30-40% slower but delivered a more thorough assessment of Ramp's marketing and sales approach. The hosts conclude that Proxy may have better browsing logic while Operator's underlying model makes stronger inferences from the collected information.

Operator took longer on the same prompt
Operator identified more detailed marketing gaps
Proxy appeared stronger or faster at browsing
Operator produced better final recommendations
The quality difference may not justify the price for every user

“my guesstimate for this specific prompt is that operator was probably like 30, 40% slower.”
Kipp · 10:30

“The output is not as good. Its ability to infer and like the underlying model around how it takes that data it got from browsing…”
Kipp · 15:30

#proxy#operator#comparison#reasoning

Takeaway15:00

Use Browser Agents When Gathering Takes Longer Than Reading

The episode's broadest use-case insight is that browser agents excel when a task requires many pieces of information from many sources. They can perform that collection in the background, leaving the user to consume and apply the consolidated output.

Multi-source collection is the core strength
Simple searches may not be sufficient
Agents can work while the user does something else
The value comes from reducing gathering time

“the gathering of the information takes way longer than the consuming of the information, right?”
Kipp · 15:00

“if you need a lot of things from a lot of different sources, these browser agents are very helpful in that”
Kipp · 15:00

#research#productivity#browser agents#automation