Topic
Ingestion, chunking, retrieval, and grounding pipelines that feed retrieved document context to an LLM.
KAUST's NoteBank lifts Gated DeltaNet's mean accuracy on six 1M-token RULER tasks from 1.7% to 70.3% with no added training. The paper's fix for state forgetting is a small archive of verbatim snippets replayed before each question.
Reality
Sayora's receptionist test found Jev gave the LLM a wrong page on 3% of 168 questions, against 33% for its embeddings. The figure is Sayora's own, from four small-business sites, and transfers only to sites built from single-topic pages.
Reality
OpenSearch Serverless NextGen, S3 Vectors and DynamoDB vector indexing now all scale to zero on AWS, where OpenSearch's old floor was about $200 a month. A developer who deployed all three from one CDK project says it is hard to look past NextGen.
Reality
Tiger Data says its pg_vectorsearch alpha answers at a 6 ms median and 95% recall on 100M vectors in its own tests. Its release notes say filtered queries can come back short of LIMIT, so we would keep pgvector HNSW in place for now.
Reality
Orca Security cut query cost 90%, input tokens 93% and latency 75% in Discovery AI, its natural-language agent for Sonar queries on Amazon Bedrock. The baseline was a one-prompt proof of concept, so the savings transfer only to teams that start from one.
Reality
Google DeepMind's EmbeddingGemma 2 maps text, images, video and audio into one vector space using 740M Apache-licensed parameters. It is built to run on phones and laptops, so the case for a hosted embedding API now depends on whether its quality holds up on a buyer's own data.
Reality
Google DeepMind released EmbeddingGemma 2, a 740M-parameter open model that runs on a phone and embeds text, code, images, video and audio in one space. Teams running a separate embedder per modality can consolidate on it if Google's reported benchmark numbers hold on their own data.
Reality
MLCommons' MLPerf Inference v6.1 adds tests that time a whole RAG pipeline and a 1,007-turn coding-agent replay. Those scores help a buyer size hardware only where the buyer's own pipeline spends its time the way the four-model reference does.
Reality
AWS says metadata pre-filtering in S3 Vectors returns up to 5x more matching vectors on highly selective filters, at no additional cost. Existing indexes keep CLASSIC search until updated, so multi-tenant RAG stores get the gain only after an index-mode switch.
Reality
Cloudflare now sells web search from Ceramic.ai, Exa and Linkup through AI Gateway at partner list prices, billed from gateway credits with no markup. Teams already on the gateway skip a separate search vendor account, though their agents still have to issue each search call.
Reality
ChatGPT is now a Very Large Online Search Engine under the EU's DSA, designated by the European Commission on Aug 31 with 159.1 million monthly EU users. Qtim CEO Anton Fokin argues that anyone shipping a web-search feature needs logs that can rebuild each answer from its sources.
Reality
Barclays expects half its developers to use Anthropic's Claude Code by the end of 2026 and most of them during 2027. Its published results so far come from a staff search tool and an email sorter, so the coding goal is still a seat count.
Reality
Plain RAG and GraphRAG got none of 31 counting and superlative questions right on a 100-question TigerGraph hackathon benchmark, an entrant reports. A COUNT from the graph, set beside the evidence actually read, shows when an answer is incomplete.
Reality
llama_index's QueryFusionRetriever writes into retriever-owned score wrappers in all four fusion modes, two more than the first patch fixed. Over a caching retriever, later queries read corrupted cached scores, and so far only an issue covers reciprocal_rerank.
Reality
Ai2 open-sourced AstaBrief 8B, which writes cited research reports in 51.1 seconds against 178.5 for Asta's Claude-powered mode. Labs that cannot send unpublished research questions to a hosted model can now run a cited-report generator on their own servers.
Perspective Coverage
3 publishersReality
One developer's test on 484 SEPA rulebook passages found that plain-English questions push several answers out of a top-5 vector search. Because the test measures each answer's rank directly, the failure shows up in retrieval, before the language model writes anything.
Reality
OpenAI retired the Assistants API on August 26, 2026, without an automatic tool to move old Threads into the Conversation objects that replace them. Teams that relied on managed threads now write that migration and set how much history each turn sends.
Reality
A developer's GraphRAG build turns Thailand's 96-section PDPA into a 190-node graph so an agent can walk from Section 26 to the fines that cite it. The published part argues the vector-search gap from vocabulary and reports no measured recall figures.
Reality
Qwen2.5-3B, wired to a local Wikipedia index, scored 52% on 150 post-cutoff questions it answers none of unaided, up from 33%, in a dev.to author's tests. Each fix targets a measured 3B failure, so a zero-shot 7B gained only 9 points from them, and the two readers' confidence intervals overlap.
Reality
Notion bought ZeroEntropy after its reranker cut latency in Notion's reranking step by 85%, by Notion's count. The hosted product is being sunset with an open-weight release, so teams that built on ZeroEntropy's API now have to decide where the model runs.
Reality
Build · September 30, 2026 · Confirmed2 publishers
Build · August 18, 2026 · Confirmed2 publishers
Build · September 22, 2026 · One report1 publisher
Build · September 18, 2026 · One report1 publisher