Wellwishers' Newsletter, July 2016
This mailing list is growing! Several investors have joined in anticipation of our next round. In July,
- we will meet angels and funds in Singapore as part of our fundraising campaign;
- we continue to develop our industry–academia research partnership; and
- we continue v2 product and v3 research and development.
Major Meetings and Relationships
In May, at Stanford’s FutureLaw conference, we connected with the U.S. LegalTech industry. Thanks to an introduction from team member Virgil, we pitched a partner at one of the top angel funds in the Valley. He was positive and indicated early interest in both our angel round and our seed round, subject to the usual Singapore-vs-US location concerns. He recommended Play Bigger, a book about category creation and domination.
Later in May, in Cambridge, MA, Meng ran into an old friend whose startup had taken money from a corporate fund in New York. They are now interested in investing also, subject to the usual Singapore-vs-US location concerns. They are getting to know us over Slack.
In June, we visited London for Tech Week. At a conference held by the Law Society, #RobotsAndLawyers, we made good connections with the LegalTech industry in the UK. We confirmed that the UK/EU market is good for the same pain point as the SG market.
On the last day of his trip, Meng was invited to a daylong ecosystem tour as part of the CFE government sub-sub-committee that he’s on, and made good connections with parts of the SG ministerial/perm-sec community responsible for startups and innovation entrepreneurship.
Long-Term Vision: Computational Legal
Play Bigger stands on the shoulders of Geoffrey Moore and Clayton Christensen. It’s about category thinking – the ultimate expression of company/market fit. Intel and CPUs; Cisco and networks; Amazon and e-commerce; Nike and sneakers.
Following the Play Bigger playbook, we have named our category “Computational Legal” with the launch of a playbill-style poster. It has already served us well in investor outreach and social media PR.
What other category kings inspire us? Adobe (design); Intuit (accounting); Autodesk (engineering). In legal, someone should be able to build an organization of similar scale in six to ten years.
Why six to ten years? Startups who turned into category kings consistently did so within that time frame. Accepting this as a law of nature, we will aim to do the same.
Why now? Because the academic research underlying contract formalization and static analysis is just maturing to the point of commercial feasibility.
Why us? Because we are domain experts in the market segment, having experience on the investor, startup, and lawyer sides; we are computer scientists who have designed and globally deployed DSLs; we have led opensource communities; we have raised venture funding and exited startups; we are well connected to distribution channels globally.
Now we take on the challenge of booting up, within a decade, an engine that spawns legal applications for many countries and many verticals. First challenge: break even on our initial product line.
Goals for Series A
We believe that product/market fit on the seed-funding app will be sufficient to justify a Series A round of about $5m. Series A will fund regional expansion for that app to Europe and the US, plus initial development of more product lines.
We already have problem/solution fit with our concierge MVP. Legalese has helped JFDI startups raise over $1.4m of funding and save an estimated $345,000. Now we want to open it up to the world, and we want the world to pay for it.
To get to product/market fit on the initial app, we need six months of product development, six months of revenue experiments, and six months of track record. Then we can start six months of fundraising.
24 months of runway before Series A.
What’s the price tag?
Our simplified projections exclude product revenues and look only at fixed costs.
Variable CAC like AdWords we aren’t going to worry about just yet. Revenues will cover those costs.
We assume each project participant is willing to work for subsistence level cash and accrue the rest of their compensation in equity.
We should have a team of 15–30 people by the end of our 24 month seed stage.
If these assumptions hold, it will cost between $1m and $2m to run for 24 months and get to Series A.
The TECS government grants reimburse $250,000 (POC) then another $500,000 (POV). So the first $750,000 that the company raises can be spent twice.
We could leverage the SPRING SEEDS program to match investor dollars 1:1.
With these multipliers, Meng could just fund the seed stage of his own pocket, but unfortunately SPRING SEEDS doesn’t match founder investment, only third-party investment. So we will go to angels and small seed funds who don’t find pre-revenue too early.
In July, we will roadshow in Singapore.
In August, we will meet investors in Silicon Valley.
In September, we will talk to investors in Cambridge and Boston.
The v1 product is mature and usable by determined early adopters, with some concierge assistance.
Revenue to date: $0.
We have started specifying and roadmapping the v2 product. We are working with developers who have previously helped to build Redmart, Mig33, and Practo. To reduce delivery risk we are planning on a blended model of in-house plus outsourced, from a vendor we have known for many years.
The recent attack on TheDAO has brought new timeliness to our DSL work. We are applying for Ethereum-related grants to help fund the static analysis work on our DSL. We will also present a series of grand challenges at Singapore’s Haskell meetup in July and hope to recruit tech volunteers there.
We are setting up a research partnership with professors and researchers at NUS, NTU, and I2R to work on the hard parts of natural language generation and static analysis.
This research is expected to reach fruition in 2018, when it will provide a significant IP differentiator relative to other players who are limited to the contract template space.
New on the team since our last letter is a young, self-taught mathematician and computer scientist. Vi says things like this:
While their CNLs seem to reliably abstract over the underlying logics, using them at all seems to compromise too many potential productivity gains. Restricting programmatic expressions to those with obvious correspondences in natural language seems inhibiting; to take an example, in Camilleri’s CL translator, the application of a bare Kleene star on an action term is non-representable because, while useful for modelling, it has no obvious NL analogue. Most troublingly, it is unclear how you’d represent abstraction over structure (think functions or subroutines) in a CNL, the absence of a solution to which would restrict “code re-use” to copy-and-paste or templating engines. For our purposes at least, I’d rather bet on education.
He’ll be helping us with the DSL and formal methods.
That’s all for this update – stay tuned for next month’s update!
written on SQ319 LHR–SIN and at teochewrestaurant.com