[ankit219]

Think this captures one of the bigger differences between what Open AI offers and what others offer using the same name. Funnily enough, Google's Gemini 2.0 Flash also has a native integration to google search[1]. They have not done it with their Thinking model. When they do we will have a good comparison.

One of the implications of OpenAI's DR is that frontier labs are more likely to train specific models for a bunch of tasks, resulting in the kind of quality wrappers will find hard to replicate. This is leading towards model + post training RL as a product, instead of keeping them separate from the final wrapper as product. Might be interesting times if the trajectory continues.

PS: There is also genspark MOA[2] which creates an indepth report on a given prompt using mixtures of agents. From what i have seen in 5-6 generations, this is very effective.

[1]: https://x.com/_philschmid/status/1896569401979081073 (i might be misunderstanding this, but this seems a native call instead of explicit)

[2]: https://www.genspark.ai/agents?type=moa_deep_research

[cxie]

deep search is the new RAG

[TeMPOraL]

Deep Search is RAG - that is, if we're still expanding the acronym instead of treating it as a word that just means "queries a vector database".

Prediction for Next Hot Thing in Q4 2025 / Q1 2026: someone will make the Nobel prize-worthy discovery that you can stuff results of your deep search into a database (vector or otherwise) and then use it to improve the ability to compile a higher-quality report from much larger amount of sources.

We'll call it DeepRAG or Retrieval Augmented Deep Research or something.

Prediction for Q2 2026: next Nobel prize awarded for realizing you may as well stop treating report generation as the core aspect of "deep research" (as it obviously makes no sense, but hey, time traveler's spoilers, sorry!), and stop at the "stuff search results into a database" and let users "chat with the search results", making the research an interactive process.

We'll call this huge scientific breakthrough "DeepRAG With Human Feedback", or "DRHF".

[hsuduebc2]

Do you find the chat interface difficult for the average user to understand? But I believe that integrating voice input—allowing users to receive spoken responses—along with a mind map that visually connects key points within the answer could make even complex information much easier to grasp.

[tough]

There's a DeepRAG paper already ser https://arxiv.org/abs/2502.01142

[TeMPOraL]

Then we'll have to call my thing Deep Embedding RAG Protocol, or DERP.