# SIG-P4B 2026-08-13 β€” Session Notes Participants: - rafa (UTC+1) - Andre Comeau | UTC-5 --- ## πŸ“– The Reading Not clearly identified in the discussion. This session was a working/scoping conversation rather than a discussion of a shared text. A prior "water project" report is referenced as a comparable prior case study, but no reading was analyzed here. ## 🧭 Overview This was a one-on-one working session between rafa and Andre Comeau to scope a potential "Protocols for Business" case study focused on the construction bidding/estimating pipeline. rafa presented findings from experiments run with his father-in-law (Virgil), a former heavy-civil construction CEO, showing that an LLM could reproduce a bid estimate within ~10% and even flag design errors from solicitation PDFs. The conversation examined why this analysis isn't already automated, identified the core bottleneck (canonical drawings live in non-machine-readable PDFs), and mapped the industry incentive structure that keeps the status quo in place. They ended by discussing how to structure the project, funding, and roles. ## πŸ’‘ Key Points & Themes - **The experiment (rafa):** rafa fed a solicitation bid PDF to an LLM (referenced as "Opus five or whatever") and, in ~30 min of runtime (~1 hour with probing/adversarial analysis and blind agent reviews), produced numbers within 10% of a full organizational bid produced by Virgil's team. He also had the model generate a "takeoff analysis." Andre confirmed "takeoff analysis" is a real, standard industry term (not just Michigan-specific). - **Takeoff analysis defined:** A table comparing the design engineer's requested materials vs. the actual planned/likely materials, and both against the contractor's read of the visual plans and real-world conditions. - **Where value really sits (rafa relaying Virgil):** The final bid number is largely deterministic (driven by subcontractor and material/equipment prices where risk is passed to others). The estimator's real leverage is in the accuracy of quantities and pricing for maximum profit opportunity. - **Andre's nuance:** Civil bids typically use unit rates, so being wrong on quantities is more forgiving than above-ground building work β€” in building, a quantity error off the drawing is "tough luck." Fixed prices (rented equipment, stone, etc.) can be plugged in with high certainty because risk shifts elsewhere; the estimator is left crunching desired margin vs. expected competitor margins. - **Finding design errors = profit (rafa):** When given a different spec, the model found a design error. This is commercially valuable because contractors often make money via change orders when they catch design errors. - **The four core questions rafa raised:** (1) Why isn't someone in the office already running PDFs through ChatGPT? (2) Why doesn't existing bidding software do this analysis? (3) Why doesn't solicitation-bid software (e.g. BidNet) do it? (4) Why is this analysis done off PDFs instead of raw CAD data? - **The root-cause diagnosis (rafa):** The real fix is upstream β€” getting machine-readable (vector) data from design firms rather than parsing PDFs. LLM PDF-parsing is viable only as an MVP / quality-control suggestion tool. rafa argued you can process a PDF faster than a $200/hr estimator to get a starting point, but you can't *remove* the estimator until you have vector-format drawings, because confidence won't be high enough. He also tied this to autonomous robot deployment: you likely can't get there without fixing the drawing/information pipeline first. - **Why CAD isn't shared (Andre):** CAD is treated as proprietary engineer IP and is hard to obtain across the industry. Engineers prefer contractors work off the drawings. Version control in construction is far weaker than in software, so there are many "cheats." CAD files often contain leftover junk/hand-overwrites; the PDF is the reviewed, "stamped," liability-bearing canonical version. Andre likened the PDF to a legacy artifact that "mimics a paper that isn't really what the future holds." - **A possible interim fix (Andre):** Reformat PDFs into machine-readable formats (e.g. markdown) β€” but the canonical drawing remains paper-format, so reading errors become the contractor's problem/margin risk. Tools like Bluebeam already do takeoffs but still read from paper-format drawings. - **The incentive/coordination problem (both):** No single party is incentivized to fix this β€” rafa framed it as a "protocol entanglement" and a "Nash equilibrium" that persists only because 100% verification between contract, design, and real-world conditions is impossible. Lawyers, insurers, engineering firms (partial liability protection), contractors (profit + risk), and owners (overpayment risk) all coexist in a "Kafka" arbitrage. Andre reframed it as **risk sharing**: reducing contractor risk reduces owner cost β€” a shared benefit β€” and disputes near the "claim / not a claim" borderline cost everyone money once lawyers get involved. - **Who should demand vs. who benefits (rafa):** The project owner is the party positioned to *demand* high-def SVG/vector exports (owners only reduce cost), while the contractor *benefits commercially* most (profit opportunity). Andre works owner-side (a government agency that owns land and does in-house design), which he noted could remove a slice of engineering risk and capture cost savings if it works β€” unlike a private firm unwilling to "put their neck on the line." - **Coordination complexity (Andre):** Modern drawings aren't from a single entity β€” a project may involve a prime architect plus structural, mechanical, and civil engineers, door/hardware and masonry specialists, etc. The prime consultant reviews all PDFs against each other and publishes the "issued for tender" set. Scaling any solution requires pushing coordination past the prime consultant into all subcontracts. - **Not a platform play (rafa):** rafa argued the right solution is *not* a new piece of software for everyone to adopt (too un-specialized), analogous to the "water project" β€” an ecosystem information-architecture problem that's "everybody's problem, but not really their problem." ## πŸ”€ Questions & Disagreements - **Is this real / does the MVP work?** rafa flagged the MVP might not pan out if LLM reliability is too low to remove the estimator. Andre said the case "looks great" but, being "heavier on the data side," he couldn't personally judge whether it's "real or not," while noting the production value/professionalism was high. - **Can you actually get the CAD/drawings?** rafa's specific ask β€” is there a "export high-def SVG/vector" option in the ~half-dozen engineering design tools? He suspected the answer should be yes today. Open question Andre agreed to check with in-house designers. - **Will firms share files?** rafa predicted an engineer would refuse to send a CAD file (embarrassed by hand-overwrites/"bullshit" in the design) as a risk/liability concern. Andre agreed the version-control/liability review is exactly why the PDF stays canonical. - **Framing difference (mild):** rafa leaned on a *profit-opportunity* framing for the contractor; Andre repeatedly nudged toward a *risk-sharing / shared-benefit* framing across owner–contractor. - **Missing collaborator:** The group is missing an engineering-firm-side "co-conspirator." Andre suggested he might be able to do this in-house (his agency has in-house engineers/architects) or put out feelers to consultants he sees on-site. Note: the two speakers frequently talked over each other (overlapping/interleaved transcript), so some exchanges are fragmentary. ## πŸ”— References Mentioned - **The "water project" / water report** β€” a prior protocol-consulting case study rafa treats as the "first rep"; this construction project would be the "second rep." Andre cited figures of ~$300k/year (salaried software management) vs. ~$5M/year California allegedly spends repeating the same study. - **Virgil** β€” rafa's father-in-law, ex-CEO (~20 years) of Dan's Excavating, a large Michigan heavy-civil contractor; ran the manual bid used for comparison. - **Dan's Excavating** β€” heavy-civil construction contractor (Michigan). - **"Opus five" (LLM)** β€” the model rafa used (as stated in the transcript). - **ChatGPT** β€” referenced as the obvious tool no one is using on these PDFs. - **BidNet** β€” solicitation-bid software where PDFs are picked up. - **Bluebeam** β€” software that can perform takeoffs from drawings. - **Protocol Institute** β€” the organization rafa proposes running the project through (funding/hiring). - **Canadian Protocol Institute** β€” Timber reportedly creating one; Andre noted a registered org would ease his agency's purchasing process. - **Venkat, Timber** β€” mentioned as having connected rafa to Andre's construction background. - **Concepts:** takeoff analysis, unit-rate bidding, change orders, "fair and reasonable review," Nash equilibrium, protocol entanglement. ## βœ… Action Items & Next Time - **rafa & Andre:** Treat this as a "Protocols for Business" case study; Andre added to the team ("I'll put you on the team for this"). - **rafa:** Work with Virgil (contractor perspective); Andre to work from the owner perspective. Still need to recruit an engineering-firm collaborator. - **rafa β†’ Virgil:** Ask Virgil to call his engineering partner (a VP) to request CAD drawings for a specific job being bid (rafa expects possible refusal). - **Andre (after vacation, ~September):** Ask in-house designers whether design software supports high-def SVG/vector export and where drawings could be made machine-readable; put out feelers to consultants seen on-site; figure out the right approval/purchasing route. - **Andre's proposed smaller stepping-stone project:** Automating the "fair and reasonable review" of contractor pricing (currently sent to consultants who can take days), which delays sign-offs on unit-bid extras/unexpected conditions. - **Funding:** rafa to draft a pitch to seek a donation to the Protocol Institute, or structure it as paid work (hire the Institute / hire rafa with a back-end donation). Registration as a formal organization (possibly the Canadian Protocol Institute) would ease Andre's agency purchasing. - **Next meeting:** rafa proposed reconnecting in a few days / "talk next week," contingent on Andre's curiosity/vacation. ## ⭐ Memorable Quotes - **rafa:** "why the fuck are you doing this analysis off of PDFs instead of the raw CAD data from the engineering department?" - **Andre:** "version control in construction is not nearly as good as version control in software." - **Andre:** "I've been working on this for thirty years almost… I'm just gonna try to profit from the chaos."