Link
Program Synthesis Pub 🍻
Search

Blogs, Newsletters & Articles đź“°

  • Alex Polozov’s “Program Synthesis in 2017–18“ – A blog post that provides a high-level overview of research progress around 2017–2018 (Program Synthesis in 2017-18 – Alex Polozov) (Program Synthesis in 2017-18 – Alex Polozov). Polozov (one of the survey co-authors) highlights emerging ideas like neural-guided search, program sketching with learned models, graph-based program representations, and new datasets. It’s essentially an update to the 2017 survey, pointing readers to about a dozen important papers and breakthroughs that came out shortly after. A great read to catch up on late-2010s trends in the field.
  • James Bornholt’s Blog – “Program Synthesis Explained” (2015) and follow-ups on his personal blog demystify synthesis for a general CS audience. With clarity and humor, he answers “what is program synthesis?” and why it matters (Program Synthesis Explained — James Bornholt). He uses simple examples to illustrate how a synthesizer figures out a program from a spec. His 2018 post “Building a Program Synthesizer” (mentioned above) is also on the blog. These posts are excellent for students or practitioners who want a gentle introduction before diving into research papers.
  • Medium: Could Program Synthesis Unlock AGI? – An essay by Ksenia Se (2023) that discusses program synthesis from an AI perspective ([#84: Could Program Synthesis Unlock AGI?by Ksenia SeMedium](https://kseniase.medium.com/84-could-program-synthesis-unlock-agi-45a43b04d4f4#:~:text=Reasoning%20Corpus%20,performance%20to%20adaptability%20and%20reasoning)) ([#84: Could Program Synthesis Unlock AGI?by Ksenia SeMedium](https://kseniase.medium.com/84-could-program-synthesis-unlock-agi-45a43b04d4f4#:~:text=exploring%20the%20fascinating%20hybrid%20of,be%20combined%20with%20deep%20learning)). It recounts François Chollet’s argument that program synthesis is key to developing more general intelligence, as it encourages AI systems to write small programs to solve tasks rather than just make predictions. The article also provides a nice historical recap ([#84: Could Program Synthesis Unlock AGI?by Ksenia SeMedium](https://kseniase.medium.com/84-could-program-synthesis-unlock-agi-45a43b04d4f4#:~:text=History%3A%20Of%20course%2C%20we%20can,trace%20Program%20Synthesis%20way%20back%E2%80%A6)) ([#84: Could Program Synthesis Unlock AGI?by Ksenia SeMedium](https://kseniase.medium.com/84-could-program-synthesis-unlock-agi-45a43b04d4f4#:~:text=,holes%20that%20are%20automatically%20filled)) and explains how the combination of deep learning and synthesis could overcome the limitations of pure neural nets. This piece is geared towards a broad tech audience, linking concepts in program synthesis to the grand goal of AGI.
  • Microsoft Research Blog – “Deep Learning for Program Synthesis” (2017) – Written by Rishabh Singh et al. (Deep Learning for Program Synthesis - Microsoft Research) (Deep Learning for Program Synthesis - Microsoft Research), this blog post describes how Microsoft was integrating deep learning with its PROSE toolkit. It gives the example of formatting strings with a few examples (Flash Fill scenario) and how a neural network can help generalize the intent (Deep Learning for Program Synthesis - Microsoft Research) (Deep Learning for Program Synthesis - Microsoft Research). It’s a concise explanation of an ML-assisted PBE system, written for a general tech readership. MSR’s blog often covers program synthesis advances (search their blog for “program synthesis” to find related posts and interviews, e.g. a profile of Sumit Gulwani’s work on programming-by-example).
  • The Morning Paper (Adrian Colyer’s blog) – This popular blog has covered a few program-synthesis papers in summary form. For instance, it featured “Optimizing SQL Queries with Program Synthesis” and “Type-Driven Program Repair” in accessible write-ups. Use the site’s search to find posts tagged with program synthesis; each provides a 5-minute read summary of a research paper, which can be easier to digest initially than the paper itself.
  • SIGPLAN Blog – Program Synthesis and Programming by Intent – The SIGPLAN blog occasionally publishes articles on emerging areas. A notable piece is a 2018 article discussing how programming by intent (through examples, natural language, etc.) is changing software development. It mentions the successes of Flash Fill and the potential of synthesis in developer tools. Such posts can give a big-picture view and often reference multiple research projects.
  • Newsletters: While there isn’t a dedicated program synthesis newsletter, related newsletters sometimes touch on it. For example, The Gradient and Import AI have included sections on code synthesis (especially around the time of Codex and AlphaCode announcements). Also, Ben Dickson’s TechTalks has explanatory articles (e.g., “AI that writes code: The future of software development?”) which heavily feature program synthesis concepts. Subscribing to general AI/PL newsletters will occasionally surface major news in program synthesis.