Askfeather.ai – Professional Class AI Tax Assistant
Hi HN, We’re the team at Feather Labs, and we built Feather (https://askfeather.ai), an AI tax assistant designed to assist how professionals handle modern Tax research. General LLMs are a liability for tax work because they lack a hierarchical understanding of the law. They often conflate IRC Title 26 with non-authoritative blog posts or outdated Treasury Regulations. We built Feather to move past "plausible" prose toward audit-defensible reasoning. The Technical Challenge: Standard RAG often chokes on the tax code for a few specific reasons: 1. Hierarchical Fragmentation: Simple…
What it does
In the maker’s words, at launch
Hi HN, We’re the team at Feather Labs, and we built Feather (https://askfeather.ai), an AI tax assistant designed to assist how professionals handle modern Tax research. General LLMs are a liability for tax work because they lack a hierarchical understanding of the law. They often conflate IRC Title 26 with non-authoritative blog posts or outdated Treasury Regulations. We built Feather to move past "plausible" prose toward audit-defensible reasoning. The Technical Challenge: Standard RAG often chokes on the tax code for a few specific reasons: 1. Hierarchical Fragmentation: Simple character-count chunking breaks the logical nesting of tax law; we implemented a strategy to preserve the relationship between code sections, sub-clauses, and court cases. 2. Temporal Decay: A vector search might pull a 2021 Revenue Ruling that was superseded in 2024; our indexing prioritizes versioning and the latest IRS guidance. 3. Contextual Overlap: The tax code is highly repetitive. We use multi-jurisdictional analysis to distinguish between federal and state-level nuances that appear semantically similar but are legally distinct. The Build: 1. Audit-Ready Citations: Every answer is grounded in primary sources—IRC Title 26, Treasury Regs, and IRS guidance, providing verified references you can actually defend. 2. Context-Aware Intelligence: Beyond answering queries, the system flags related risks, exceptions, and filing deadlines that practitioners might overlook. 3. Compliance: We are SOC 2 compliant with a strict zero-training policy on sensitive client data. We’re curious to hear from anyone else building RAG for "dense" domains where "close enough" results are not enough. We'll be in the comments to talk about our indexing strategy and how we handle complex document extraction. Sokratis
Does the same job
all alternatives →
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 27d ago · cactuscompute.com
