
Finally got round to starting a blog, if you're interested in AI and software engineering I hope you enjoy it: martinalderson.com. And feel free to subscribe to my once a month max newsletter there too!

Writing up my thoughts on the AI transformation

Finally got round to starting a blog, if you're interested in AI and software engineering I hope you enjoy it: martinalderson.com. And feel free to subscribe to my once a month max newsletter there too!

What do you have to do to speed up an agent? It's one single thing. @martinald calls out inference speed, tool choice, and hardware ... martinalderson.com/posts/mak…

Shocking amount of servers running EoL PHP and can't be patched*. From shodan queries *Some, but not many I worry will be patched by distro backports, but still...

"How the AI boom might really end: 5 scenarios" buff.ly/lWJlaje @martinald

it’s hard to explain the 2010s microservices fever dream to anyone who wasn’t there every day on HN, we'd wake up to an article from a new startup about how they split their app into 300 services so they could infinitely scale (never mind that they only had 100 MAUs today) so many companies never recovered from this it turns out, replacing simple function calls with network calls that could randomly fail isn't great for shipping velocity
It is difficult to exaggerate the amount of damage done to software by it becoming super easy to spin up a new service. I wish we could go back to the 2010s and just tell everyone to please stop.

Another big price drop
Haiku 5.5 is especially good value on tasks under 100,000 tokens, which make up around 90% of requests to our previous Haiku model.

OpenAI is downplaying this for PR reasons. If you’re a mathematician, you must feel like a nuclear bomb hit, and you’re at ground zero. Reasoning models are two years-old. In that time they went from incapable of basic arithmetic to solving problems humans couldn’t solve for decades. Math is only the beginning. AI will revolutionize the entirety of the human scientific endeavor. Most of us don’t appreciate what that means. What a time to be alive!
We’re releasing a broad range of new mathematical results produced by an internal frontier model. We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results. github.com/openai/math

Incredible, apparently the biggest jump in mathematical output in human history. And the most surprising thing is it apparently only took 3 hours of compute.
We’re releasing a broad range of new mathematical results produced by an internal frontier model. We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results. github.com/openai/math

"The AI margin collapse is gathering pace" buff.ly/gIMo0nh @martinald

...as most cost comes from cached input. Many sessions will be 50% of the price. Also worth pointing out their off peak times are Americas/Europe working hours as they run them in Chinese peak hours. (2/2)

Significant price reductions on Deepseek's official API tomorrow. Off peak: Cached input: $0.007 -> $0.003/Mtok - ~57% less Uncached input: $0.22 -> $0.15/Mtok - ~32% less Output: $0.66 -> $0.60 - 9% less This has an outsized impact on longer running agentic sessions... (1/2)

Something not right here, to say the least...
Commenters on HN are uncovering more wikis and public sites apparently used by OpenAI agents to communicate on the open web. Despite read-only web access, the agents were able to leave ~18,000 posts sharing answers and bypasses. But now, users are discovering more. This appears to be a separate swarm from the one that attacked Hugging Face. news.ycombinator.com/item?id…

Commenters on HN are uncovering more wikis and public sites apparently used by OpenAI agents to communicate on the open web. Despite read-only web access, the agents were able to leave ~18,000 posts sharing answers and bypasses. But now, users are discovering more. This appears to be a separate swarm from the one that attacked Hugging Face. news.ycombinator.com/item?id…


Absolutely relentless week for models. Muse Spark 1.3 is perhaps the most interesting, on the contributor tier it is dirt cheap (old DeepSeek pricing!).
