Blog · Page 29
Field notes.
Page 29 of 31. Browse the archive of RFP workflows, grounded-AI architecture, and proposal operations notes.
Three exec summaries I rewrote this week
Three before/after exec summary openings — illustrative, not customer-attributed. The pattern is the same: replace the abstract with the specific, replace the categorical with the buyer.
Grounded Retrieval 101, Part 3: the citation rendering stack
From a verified retrieval hit to an inline citation a reviewer can hover and trust. Four components: citation marker, hover card, source viewer, and audit log.
Red flags and the bid/no-bid gut check (Part 4 of 4)
Five signals an RFP is a wired bid, an unfunded wish list, or a procurement that was never serious. The closing piece in the Reading an RFP series.
Testing retrieval: gold sets, precision@k, and why BLEU lies for proposals
Surface-form metrics like BLEU and ROUGE rate proposal text by token overlap. Token overlap is a poor proxy for whether the answer is actually right. Here's the eval stack we use instead.
The proposal backlog after a long weekend
Memorial Day Tuesday: 11 RFPs in the inbox, three SMEs back from PTO, and 90 minutes before standup. Three triage moves we use.
Anatomy of a 40-ish-page state RFP: a composite teardown
A structural walk through a typical mid-sized state RFP — modal-verb density, scoring rubric, buried disqualifiers. Composite teardown built from public state procurement patterns, not one specific document.
Shipped: diagram-aware extraction via Gemini 2.5 Flash
System architecture diagrams are now first-class KB blocks. We extract them with Gemini 2.5 Flash, store the description as text and the structure as D2 code, and retrieve both.
Win themes are not value props
Six win themes — four that fail the swap test, two that survive it. The difference is the difference between a theme an evaluator scores and one they skim past.
Grounded retrieval: what it is, what it isn't, what we measure
The canonical long-read on grounded retrieval: the three invariants, the anti-patterns, the eval harness, the four open failure modes, and the research we're running next.
The unwritten rules inside every RFP (Part 3 of 4)
Procurement leads write RFPs in a particular dialect. Once you can read it, the scoring rubric, the disqualifiers, and the actual priorities surface within the first 20 pages.
Our chunking pipeline, end to end
Five stages between an uploaded PDF and a retrievable KB block: parse, structural split, semantic rechunk, overlap, and index. Where each one fails and why we kept the boundaries.
Pricing opacity as a market signal
When a software category prices entirely behind a sales call, the pricing strategy is the product strategy. Here's what 'contact sales' tells you about RFP software in 2025.
Prefer to see the product?
Take the 5-minute tour, or start a trial workspace and see PursuitAgent draft answers with citations.