--- schema: 1 type: video title: I Spent Two Years Building the Median — Hariprasad Sudharshan description: >- After two years of intense solo building, Hariprasad Sudharshan discovers that other people have independently produced calculators much like his own. Nobody needs to have copied anyone: the same models, tools and familiar problem can lead builders toward the same solution. He calls that shared starting point the median. Faster development makes it easier to ship useful software, including internal tools and learning projects, but a functioning demo does not establish a durable product advantage. His talk is a personal reflection on that distinction, rather than a controlled study of product success or a disclosed business outcome. Sudharshan separates the model's capability from its harness and the product a customer ultimately receives. He describes a build first reflex that replaces searching for existing solutions with immediately opening a coding agent. His alternative is a novelty budget: name the one distinctive thing before starting, include it in the roadmap and review it as deliberately as other project spending. For leaders evaluating demos, he proposes three questions. Who has already solved this? What does this team have that a competitor does not? Where will the application reach users? A later calculator iteration focuses on the difference between running open model weights inside an enterprise and purchasing inference, including utilization and the infrastructure between weights and generated tokens. The practical lesson is to use the readily available starting point where it helps, then spend judgment on a specific problem, distribution and a defensible difference. Speaker info: - https://github.com/brettleehari - https://www.linkedin.com/in/haripm4ai/ Timestamps: 00:00 - The deterministic software underneath everyday life 3:54 - Rented models change how software gets built 6:41 - Athletics and moving model leaderboards 10:28 - Model, harness and product 12:09 - The median and converging solutions 14:24 - Useful software and the limits of differentiation 18:03 - The build first reflex 20:39 - A calculator and two years of solo building 23:51 - Define a novelty budget 26:19 - Four moves for a product roadmap 28:30 - Three questions for leaders evaluating demos 30:45 - Open weights, inference costs and the revised calculator 33:42 - Spend judgment on a distinctive problem canonical: >- https://liquidgold.cc/channel/aidotengineer/explore/video/i-spent-two-years-building-the-median-hariprasad-su-2db44d434361 markdown: >- https://liquidgold.cc/channel/aidotengineer/video/i-spent-two-years-building-the-median-hariprasad-su-2db44d434361.md llms: null parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-23T22:23:57.071Z' language: null children: [] videoId: 0JZy5AuuQzc durationSeconds: 2171 publishedAt: '2026-10-14T18:00:11.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 13 endSeconds: 236 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=13' heldReason: rights - ordinal: 1 text: null startSeconds: 13 endSeconds: 236 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=13' heldReason: rights - ordinal: 2 text: null startSeconds: 13 endSeconds: 236 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=13' heldReason: rights - ordinal: 3 text: null startSeconds: 247 endSeconds: 411 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=247' heldReason: rights - ordinal: 4 text: null startSeconds: 247 endSeconds: 411 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=247' heldReason: rights - ordinal: 5 text: null startSeconds: 422 endSeconds: 628 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=422' heldReason: rights - ordinal: 6 text: null startSeconds: 422 endSeconds: 628 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=422' heldReason: rights - ordinal: 7 text: null startSeconds: 641 endSeconds: 730 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=641' heldReason: rights - ordinal: 8 text: null startSeconds: 641 endSeconds: 730 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=641' heldReason: rights - ordinal: 9 text: null startSeconds: 740 endSeconds: 864 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=740' heldReason: rights - ordinal: 10 text: null startSeconds: 876 endSeconds: 1085 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=876' heldReason: rights - ordinal: 11 text: null startSeconds: 876 endSeconds: 1085 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=876' heldReason: rights - ordinal: 12 text: null startSeconds: 1097 endSeconds: 1251 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=1097' heldReason: rights - ordinal: 13 text: null startSeconds: 1261 endSeconds: 1431 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=1261' heldReason: rights - ordinal: 14 text: null startSeconds: 1443 endSeconds: 1586 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=1443' heldReason: rights - ordinal: 15 text: null startSeconds: 1597 endSeconds: 1720 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=1597' heldReason: rights - ordinal: 16 text: null startSeconds: 1597 endSeconds: 1720 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=1597' heldReason: rights - ordinal: 17 text: null startSeconds: 1731 endSeconds: 1847 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=1731' heldReason: rights - ordinal: 18 text: null startSeconds: 1731 endSeconds: 1847 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=1731' heldReason: rights - ordinal: 19 text: null startSeconds: 1857 endSeconds: 2023 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=1857' heldReason: rights - ordinal: 20 text: null startSeconds: 1857 endSeconds: 2023 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=1857' heldReason: rights - ordinal: 21 text: null startSeconds: 2034 endSeconds: 2171 watch: 'https://liquidgold.cc/watch/0JZy5AuuQzc?t=2034' heldReason: rights --- # I Spent Two Years Building the Median — Hariprasad Sudharshan After two years of intense solo building, Hariprasad Sudharshan discovers that other people have independently produced calculators much like his own. Nobody needs to have copied anyone: the same models, tools and familiar problem can lead builders toward the same solution. He calls that shared starting point the median. Faster development makes it easier to ship useful software, including internal tools and learning projects, but a functioning demo does not establish a durable product advantage. His talk is a personal reflection on that distinction, rather than a controlled study of product success or a disclosed business outcome. Sudharshan separates the model's capability from its harness and the product a customer ultimately receives. He describes a build first reflex that replaces searching for existing solutions with immediately opening a coding agent. His alternative is a novelty budget: name the one distinctive thing before starting, include it in the roadmap and review it as deliberately as other project spending. For leaders evaluating demos, he proposes three questions. Who has already solved this? What does this team have that a competitor does not? Where will the application reach users? A later calculator iteration focuses on the difference between running open model weights inside an enterprise and purchasing inference, including utilization and the infrastructure between weights and generated tokens. The practical lesson is to use the readily available starting point where it helps, then spend judgment on a specific problem, distribution and a defensible difference. Speaker info: - https://github.com/brettleehari - https://www.linkedin.com/in/haripm4ai/ Timestamps: 00:00 - The deterministic software underneath everyday life 3:54 - Rented models change how software gets built 6:41 - Athletics and moving model leaderboards 10:28 - Model, harness and product 12:09 - The median and converging solutions 14:24 - Useful software and the limits of differentiation 18:03 - The build first reflex 20:39 - A calculator and two years of solo building 23:51 - Define a novelty budget 26:19 - Four moves for a product roadmap 28:30 - Three questions for leaders evaluating demos 30:45 - Open weights, inference costs and the revised calculator 33:42 - Spend judgment on a distinctive problem Canonical: https://liquidgold.cc/channel/aidotengineer/explore/video/i-spent-two-years-building-the-median-hariprasad-su-2db44d434361 · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md) > Search chunk text is held until public-source provenance is verified. ## Document chunks --- #### Chunk 0: Document excerpt (0:01 to 3:56) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 1: Document excerpt (3:56 to 6:51) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 2: Document excerpt (6:51 to 10:28) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 3: Document excerpt (10:28 to 12:10) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 4: Document excerpt (12:10 to 14:24) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 5: Document excerpt (14:24 to 18:05) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 6: Document excerpt (18:05 to 20:51) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 7: Document excerpt (20:51 to 23:51) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 8: Document excerpt (23:51 to 26:26) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 9: Document excerpt (26:26 to 28:40) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 10: Document excerpt (28:40 to 30:47) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 11: Document excerpt (30:47 to 33:43) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 12: Document excerpt (33:43 to 36:11) > Held: transcript rights do not permit this chunk's text.