Hacker News: sanjitb

New comment by sanjitb in "Early experiments in accelerating science with GPT-5"

sanjitb — Sun, 23 Nov 2025 18:00:43 +0000

pretty cool stuff!

Early experiments in accelerating science with GPT-5

sanjitb — Sun, 23 Nov 2025 18:00:43 +0000

Article URL: https://openai.com/index/accelerating-science-gpt-5/

Comments URL: https://news.ycombinator.com/item?id=46025776

Points: 4

# Comments: 1

New comment by sanjitb in "Cloudflare outage on November 18, 2025 post mortem"

sanjitb — Wed, 19 Nov 2025 05:34:12 +0000

thanks for clarifying! i guess then they never explained why the status page went down, even though it's supposed to be running on independent infrastructure.

New comment by sanjitb in "Cloudflare outage on November 18, 2025 post mortem"

sanjitb — Wed, 19 Nov 2025 04:49:38 +0000

cloudflare:

> Throwing us off and making us believe this might have been an attack was another apparent symptom we observed: Cloudflare’s status page went down. The status page is hosted completely off Cloudflare’s infrastructure with no dependencies on Cloudflare.

also cloudflare:

> The Cloudflare Dashboard was also impacted due to both Workers KV being used internally and Cloudflare Turnstile being deployed as part of our login flow.

New comment by sanjitb in "Gemini with Deep Think achieves gold-medal standard at the IMO"

sanjitb — Mon, 21 Jul 2025 20:46:36 +0000

Why do they brag about not using a theorem prover? To me, whatever tool helps the model perform, go for it.

Besides, they still specialized Gemini for the IMO in other ways:

> we additionally trained this version of Gemini on novel reinforcement learning techniques that can leverage more multi-step reasoning, problem-solving and theorem-proving data. We also provided Gemini with access to a curated corpus of high-quality solutions to mathematics problems, and added some general hints and tips on how to approach IMO problems to its instructions.

New comment by sanjitb in "Expanding on what we missed with sycophancy"

sanjitb — Fri, 02 May 2025 19:46:02 +0000

> the update introduced an additional reward signal based on user feedback—thumbs-up and thumbs-down data from ChatGPT. This signal is often useful; a thumbs-down usually means something went wrong.

> We also made communication errors. Because we expected this to be a fairly subtle update, we didn't proactively announce it.

that doesn't sound like a "subtle" update to me. also, why is "subtle" the metric here? i'm not even sure what it means in this context.

New comment by sanjitb in "Doing laundry on campus without a phone"

sanjitb — Sat, 02 Sep 2023 16:25:37 +0000

Speaking of the evils of internet-reliant devices and bad engineering: those MIT washing machines actually went down for two weeks. MIT changed the authentication configuration of their wireless network, the software couldn't re-connect, and all hell broke loose. They had to turn on some emergency operating mode that allowed students to use them for free, without any app.