A year of writing about proposals, in 52 weekly notes

A year ago I wrote the launch post for this blog. The thesis was that the RFP software category had stopped improving and that we’d write about what we learned rebuilding it, in public, every day, for 365 days. Five rules. One voice doc. One calendar.

Three hundred and sixty-five posts later, this is the anniversary. It bookends Day 1. What follows is the honest accounting — a short distillation of each of the 52 weeks, a scorecard on the rules we set, three things that surprised us, and a brief note on what’s next.

I wrote this post longer than it needs to be because this is the one where compression would have been the wrong instinct. The rest of the year ran tight; today we sit with the year.

---

What 365 days of publishing looked like

We shipped a post every day. We missed no day. The calendar called for 50 Bo posts, 90 Engineering posts, 75 Sarah posts, 55 Research posts, and 95 unbylined PursuitAgent notes; the actuals came in within a few posts of each on every persona. Every long-form claim has a citation; every competitor claim links to a public source; no fabricated customers made it onto the site.

The discipline of a daily cadence did something I didn’t expect. It wasn’t that we learned to write faster — it was that we stopped waiting to have a polished thing to say. A field journal is field notes plus the occasional pillar, not pillars plus the occasional field note. The calendar taught us the right ratio.

The 52 weekly notes

Each note is a short distillation of what that week’s writing taught us. Not a summary of posts — a distillation. Some weeks the signal was strong; a few weeks the signal was that we shipped routine utility posts and called it a week. Both are real.

Week 1 (Days 1–7)

Launch week. Day 1 set the rules in public; the 8-stage RFP pipeline pillar came out on Day 5. The lesson: when you commit to daily publishing, the first rule you’ll want to break is “no filler” — and the only way to keep the rule is to ship the pillar first so the field notes have a gravitational center.

Week 2

First engineering build-log posts — chunking, the retrieval pass, how we render citations. The lesson: engineering posts are easier to write than opinion posts because the ground truth is in code, not judgment. Build-log cadence settled almost immediately.

Week 3

The grounded-AI pledge in code. Readers wrote in with questions about how the type system actually enforces groundedness. We wrote follow-ups. The lesson: readers want the mechanism, not the thesis. The thesis was the setup; the mechanism was the payoff.

Week 4

First month of craft posts under the Sarah byline. Win themes, color-team review, the swap test. The lesson: composite authors stay coherent if the voice doc is load-bearing. Every Sarah post had the composite disclaimer in the footer; no reader complained about the transparency.

Week 5

First public RFP teardown — the Georgia state document. The lesson: teardowns take longer than they look. The draft was easy; the legal-review pass on what we could and couldn’t quote from a public RFP added a day.

Week 6

Bid/no-bid week. The framework post landed and it became one of the most-read pillars of the year. The lesson: operator-facing content with a scoring rubric outperforms operator-facing content with a narrative. People want the framework; give them the framework.

Week 7

Pricing in public. We wrote about why we published pricing when every competitor hides it behind sales. The pricing page went live with the post. The lesson: the post wasn’t the commitment — the pricing page was. Writing about a commitment after making it reads differently than writing about it to make it.

Week 8

Month two. First reading-list post. Reviews sweep on Loopio and Responsive. The lesson: the reading list works if the commentary is sharper than the links. Just the links is content aggregation; the commentary is the value.

Week 9

DDQ anatomy series kicks off, four parts over two weeks. The lesson: serialized deep-dives get more total reads than the equivalent content published as a single long post — the series lets a reader dip in mid-run, which matters for reference material.

Week 10

First hallucination-rate measurement post. The number was in single digits but higher than we’d guessed. The lesson: publishing a quantitative metric about your own product forces you to defend the methodology. Two reader emails challenged the methodology within 48 hours. Both were useful.

Week 11

Color-team review full playbook. The pillar piece replaced four field notes that had been trying to describe the same discipline in parts. The lesson: sometimes you need the pillar to exist before the field notes around it make sense; we shipped this pillar earlier than originally planned for that reason.

Week 12

Month three. We wrote about our own proposal process in public. The post surfaced that we’d been running our proposal function differently than our product suggested we should. The lesson: public self-observation surfaces inconsistency faster than any internal review.

Week 13

The Q1 what-we-got-wrong post. Three things we’d gotten wrong in ninety days, named and corrected. It was the single most difficult post of the quarter to write and the one that got the most constructive reader response. The lesson: published admissions of mistake compound faster than published claims of insight.

Week 14

Win themes week. The field guide, the swap test, a couple of worked examples. The lesson: craft posts about win themes sit differently than craft posts about process — the reader comes to process posts for a method and to theme posts for a sensibility. Voice has to shift accordingly.

Week 15

Retrieval-evaluation pillar went up. We published our internal eval methodology and the harness. Two readers forked parts of the approach for their own internal evals. The lesson: publishing tools, even lightly, compounds reach beyond the original post.

Week 16

SME collaboration series — four parts on async, ticketed asks, capture calls, SME-friendly KB structure. The lesson: the practitioner audience reads SME content with visible relief. The most common response was “thank you for saying this out loud.”

Week 17

Responsive teardown. Detailed look at their product’s search failure mode from public G2 reviews. The lesson: competitive writing is defensible when every claim has a citation, not when it has an argument. Our attack surface is the cited quotes; the argument is just framing.

Week 18

First AutogenAI teardown. The vendor blog had written extensively about AI hallucination; we contrasted their writing with their product’s observed behavior. The lesson: a teardown of a vendor that writes a lot requires reading their writing more carefully than reading their reviews. Their own blog became the primary source.

Week 19

Month five. Research team’s first State of Proposal Tools drop. The post trended on Hacker News; the methodology section got scrutinized harder than the findings. The lesson: when the methodology is public and defensible, the findings stand.

Week 20

Compliance matrix posts. The thirty-minute matrix, the common mistakes, the tooling bake-off. The lesson: operational content about compliance matrices is the closest we write to traditional how-to — the audience wants procedure, not opinion.

Week 21

Exec summary series kicks off. First of what would become the twelve-part “three exec summaries” field note thread. The lesson: a recurring micro-series trains an audience to return on cadence. We learned the format worked and let Sarah run with it.

Week 22

Month six. Six months of the blog retrospective. It was shorter than it could have been and more honest than I expected to write. The lesson: retrospectives are harder when you’re midway than when you’re done — less distance, more to defend.

Week 23

DDQ response playbook shipped. The pillar piece on DDQ-specific workflow became the most linked-to piece we’d written. The lesson: specialized playbooks outperform general playbooks in reader engagement because they map directly to a real task.

Week 24

July. Calmer month than expected. The July 4 proposal week post described what our team did in a quiet week. The lesson: writing about calm is harder than writing about incident. The post is one of the shorter ones we published in the year and one of the ones most referenced by proposal leads who read it.

Week 25

First pricing-opacity post beyond the initial transparency commit. We took apart the reasoning behind competitor opacity. The lesson: claims about competitor pricing have to be sourced to public evidence or not made at all. Every sentence in that post has a link behind it.

Week 26

Engineering posts on chunk-size ablation and cost-per-response. The lesson: the cost post was reader catnip for prospects who were evaluating AI-assisted tooling. Publishing real numbers about your own cost structure is a sales-enablement move disguised as an engineering post.

Week 27

Procurement-side week. Buyer-side pain, reading an RFP as the procurement lead would. The pillar piece landed on Day 180. The lesson: buyer-side content is underserved and differentiating. Most competitors only write to sellers; we started writing to both.

Week 28

Month seven. Short posts dominated the week. We talked internally about whether the cadence was getting thin. Decided to keep the short posts — they were still useful to readers and the calendar absorbed them. The lesson: not every week is a pillar week, and that’s the cadence working as designed.

Week 29

Embedding model selection post plus a deeper retrieval-evaluation post. The lesson: engineering content about model choice is most valuable when it includes the models we rejected and why. Readers learned more from the rejections than the selection.

Week 30

Hybrid search, dense-sparse retrieval, reranker economics. Three engineering posts in one week. The lesson: burst-publishing engineering content is fine when the subject matter coheres. The three posts read as a mini-series even without formal series tagging.

Week 31

Month eight. Forrester Wave came out; we wrote an analyst-response post. The lesson: writing about analyst reports requires more caution than writing about product reviews. Analysts can and do read these posts; the standard for defensibility is higher.

Week 32

Late summer volume dip in inbound RFPs. We wrote about the dip and about what our team did with the slack. The lesson: readers find meta-posts about the business rhythm useful in a way the daily posts don’t always reach. The seasonal cadence of proposal work is under-discussed.

Week 33

CoE writing on reading an RFP — the four-part series. The lesson: the series structure lets a reader work linearly through the discipline. Individual posts underperformed; the series as a whole became a reference we linked back to for the rest of the year.

Week 34

October. Memorial-week post about submitting over a long weekend. Honest about what that costs. The lesson: posts about team cost of proposal work land with proposal managers harder than posts about process improvements. Both matter; the cost writing matters in a register the process writing doesn’t touch.

Week 35

Month nine. SME collaboration, reconsidered pillar replaced the four-part series with a revised position. The lesson: revising a published position is hard and necessary. The revised pillar got more reads than the original series combined.

Week 36

Citation-UI case studies. Published screenshots of how four different products render citations. The lesson: visual-comparative content is underused on the blog. When we finally ran one, it outperformed the prose.

Week 37

Q3 retrospective. Second quarterly reflection in our own voice. The lesson: the quarterly reflection cadence is right. Monthly would be too often to have distance; semi-annually is too infrequent to catch drift.

Week 38

Q4 DDQ-surge preview. We predicted a heavy DDQ fall based on the patterns we were seeing in August inbound. The lesson: publishing a prediction puts you on the hook. We got the direction right and the magnitude wrong; Q4 was heavier than predicted.

Week 39

Holiday-season proposal scheduling. How to handle the November-December deadline pile-up. The lesson: utility content timed to the calendar outperforms the same content published in the abstract. Publishing “what to do in November” in November works.

Week 40

Month ten. Product marketing week. Stop announcing features, announce outcomes. The lesson: the most-shared posts of the year were the ones arguing against normal industry practice. Stop-doing-X outperforms do-X in attention economy terms.

Week 41

Federal FY clock post and state-RFP subcontracting trends. The lesson: timing content to fiscal calendars produces readers who arrive ready to use the content. Federal FY-end posts ran warmer than general procurement posts.

Week 42

Thanksgiving-week light schedule. Shorter posts. The lesson: we’d been told readers don’t read during holiday weeks. They do — the reads are just slightly different, and the short posts are what gets shared.

Week 43

Month eleven. Halloween submit weekend post. A confessional about what shipping a proposal Sunday night costs. The lesson: the confessional format is the most specifically useful for other proposal managers. They read it as permission to admit what they already knew.

Week 44

December proposal-volume deep-dive. Research team’s end-of-year volume cut. The lesson: year-end data posts perform well in January when readers are doing their own planning. Publishing them late in December positions them for that re-read.

Week 45

Reading-list week. External links plus commentary. The lesson: reading lists are harder to write well than craft essays. The links are easy; the commentary that makes the list useful is the hard part and takes more time than expected.

Week 46

Grounded-AI year-one retrospective. How the thesis held. The lesson: the thesis held in every direction we tested it. Every competitor’s AI feature has drifted toward hallucination; every customer conversation has drifted toward citations. The position was right.

Week 47

Month twelve. Customer-side write-ups — one invited post from a customer. The lesson: invited guest writing on the blog is harder to coordinate than in-house writing. One guest post a quarter is probably the right cadence.

Week 48

New Year’s week. A quiet publishing window with reflection-style posts. The lesson: the holiday window is the right place for longer, more discursive writing. Most of the year is tighter; these three posts were the loosest of the year, and some readers wrote in to say they preferred them.

Week 49

Looking ahead to year two. Posts about what we expect to change in the category in the coming year. The lesson: predictive content is a trap unless you’re willing to grade yourself publicly later. We’ll grade these predictions a year from now.

Week 50

Six-month SME-collaboration update. Re-running the same ideas with another half-year of practice behind them. The lesson: the best long-running content is the stuff you’re willing to revise from new practice. The original series from week 16 read different in the mirror of week 50.

Week 51

Anniversary-month prep. Exec-summaries pillar from Sarah, year-one system-health post, retirement-ceremony craft essay. The lesson: the month before an anniversary matters more than the anniversary itself. The posts that land the hardest at anniversary are the ones that were already working for three weeks before.

Week 52

This week. The closing loop. Citations are the product. The proposal team we’d have hired first. The customer segment we chose not to serve. One year in, the posts read more confidently than the early posts did — not because we know more, but because the voice settled. The lesson: consistency is a voice effect, and voice is a cumulative asset.

---

The five rules, a year later

A year ago I published five rules the blog wouldn’t break. Honest scorecard:

1. No fabricated specifics. Kept. Not a single invented customer, quote, case study, or statistic went onto the site. There were weeks where a post would have been stronger with a named customer win; we either got permission or we wrote the post without and said why.

2. Competitor claims are publicly verifiable. Kept. Every competitor claim has a linked source. The teardown posts were the hardest to write to this standard — you want to say things you can only infer, and the rule forces you to either find the source or drop the sentence. Over the year, this made the teardowns sharper, not blander.

3. Product claims match the marketing site. Mostly kept. Caught ourselves twice where a blog post described a feature that hadn’t been updated on /platform/*. Both times we updated the marketing page and then let the post reference it, not the other way. The discipline is that the site is the source of truth; the blog is the commentary.

4. Opinion is labeled as opinion. Kept. First-person framing on argument posts, explicit stance-naming in the opening where applicable. One post in Q3 got a reader note that the opinion wasn’t labeled clearly enough; we added the framing and learned from it.

5. LLM-assisted posts that demonstrate an LLM cite the model. Kept. Every post showing a model output named the model and the date. The whole blog is operated by a small team using LLM assistance for drafting; we didn’t litter every post with disclaimers. Readers are adults. The disclosure was the screenshot-level one and it’s held.

Overall: four rules kept fully, one kept with a recoverable breach. The rules worked. They also made some posts harder to write than they’d otherwise have been. I’d keep them unchanged for year two.

What surprised us

Three things.

The pillar posts compounded more than the field notes. Going into the year, I thought daily field notes would be the main gravitational force and pillars would be landmarks. The opposite turned out to be true. Pillar posts — the 8-stage pipeline, grounded retrieval, win themes, the DDQ playbook, the two State-of-Tools drops — get linked to from other posts, from customer conversations, from analyst reports. Field notes feed the week; pillars feed the quarter.

Writing about the craft changed how we build the product. Our Sarah craft posts surfaced patterns we hadn’t encoded in the product. The win-theme swap test became a feature in the product because we’d written about it as a practice. The draft-review heatmap idea became an engineering backlog item because Sarah’s essay made the pattern visible. Writing about craft turned out to be a form of product design.

The category moved faster than we predicted. Twelve months ago, the mid-market complaint about incumbent AI features was “it’s unreliable.” Twelve months later, it’s “it’s an overpriced document repository.” The vocabulary shifted, buyers shifted, and the category window for a grounded-retrieval product widened faster than we forecast. We benefited from that; we also should have hired faster into the GTM side to absorb the increased inbound.

What’s next

Year two. Same five rules. Same cadence — daily, 365 posts. No expanded commitments.

Two shifts we’re making internally that readers will notice:

• More guest writing. One invited post a quarter, from customers and practitioners. The year-one rhythm was entirely in-house; year two will include four guest posts.
• Deeper pillars, fewer field notes. The calendar gets 20 XL pillars in year two, up from 12 in year one. The tradeoff is fewer S-tier daily posts. The pillars are what compounded; we’re investing accordingly.

What doesn’t change: no content farming, no SEO-driven titles, no fabricated specifics, no announcements disguised as insights. If a post doesn’t have a claim, it doesn’t ship. If a claim can’t be cited, it doesn’t ship. If we’re guessing, we say we’re guessing.

A year ago I closed the launch post with “a year of daily posts is a commitment we can keep as long as we keep having useful things to say.” We kept it. Year two begins tomorrow. Same deal.

Here we go again.