Jamie Brandon

Tinkering with systems.

Occasionally available for consulting.

Highlights:

Reflections on a decade of coding - what has and hasn't worked for me.
Internal consistency - why most streaming systems can produce unboundedly wrong outputs.
Against SQL - SQL is inexpressive, incompressible, non-porous and under-specified, and this creates drag on everything downstream.
The shape of data - taking notation seriously.
Have you tried rubbing a database on it? A day of lightning talks about turning a data-centric lens onto familiar problems to yield strange new solutions (and maybe exciting new problems).
Pain we forgot - teaching everyone to code is not enough.

Programming language experiments:

droplet
- ~2012
- Goal: concisely specify distributed systems.
- Loosely based on bloom.
- Didn't have the skills to complete it.
eve
- 2013 Dec - 2016 Jan (the others continued until 2018 Jan)
- Goal: 'end-user programming'.
- Shipped many demos
- Lack of concrete questions/problems.
- Plagued by performance problems.
imp v0
- 2015 Aug - 2016 Oct
- Goal: Write a webapp end-to-end in a relational language.
- A practical query compiler in 500 lines
- Relational UI
imp v1
- 2018 Apr - 2020 Mar
- Goal: 'A better SQL' (see Against SQL)
- Features:
  - Denotation + type system fit on a single page.
  - Compiles to relational algebra.
  - First-class/nested sets.
  - Functions are infinite sets, function application is a join.
  - Subqueries are a join against an anonymous function.
- Core language
- Denotational semantics
- Types
- Simple interpreter
- Decorrelation
- Solving functions
- Boxes
imp v2
- 2020 May - 2022 Jan
- Goal: Push v1 towards being a capable general-purpose language.
- Same core language as v1, new runtime.
- Improved type inference and error reporting.
- First-class transactions.
- Iteration
- Implicit ordering in relational languages
- Making live repls behave
- Live repl demo
- Heterogenous types and gradual typing
imp v3 / preimp
- 2022 Jun - 2022 Aug
- Goal: Focus on the DX for state and UI.
- Draft: The program is the database is the interface.
- Dev teasers in 0022, 0025, 0026, 0027.
- Abandoned to work on TigerBeetle.
zest

Streaming/incremental systems:

An opinionated map of incremental and streaming systems
Internal consistency - why most streaming systems can produce unboundedly wrong outputs.
Why query planning is hard
Thoughts on benchmarking
dida - differential dataflow for mere mortals
- Why isn't differential dataflow more popular?
- How dida works
- Roughly works, but definitely not production-ready. Likely to be replaced by something simpler (eg).

Databases:

Interfaces:

focus - a low-latency text editor which I use for all my writing and coding
- Intro
- Rendering
- Text
reltron - an interface for exploring relational databases, in collaboration with Kevin Lynagh
monolog - monolog is to logging as spreadsheets are to accounting?
unnamed - experimenting with VR game mechanics on the HTC Vive

Programming languages:

Distributed systems:

Causal ordering
msc thesis - tweaking the timing model of a gossip algorithm to collapse the size of the state space and allow model checking fairness properties

Practices:

Community:

Rambling:

Cogsci:

Misc: