Lepore

philosophy beginner

[Zvi Mowshowitz @TheZvi]

I notice that there is rather obviously no reason to be holding back GPT-5.6-Luna, it is clearly not an issue, yet here we are.

[Nous Research @NousResearch]

Hermes Agent can now /learn from anything: feed it directories of any source material (code, API docs, manuals, PDFs, configs) and it distills a verifiable reusable skill

[NotebookLM @NotebookLM]

Flashcards are now fully customizable. Edit questions, tweak answers, and add brand-new cards to create the ultimate set of study tools.

Share them with your friends, classmates, academic rivals, etc— just be prepared for the endless shower of praise and accolades that follows.

[Massachusetts Institute of Technology (MIT) @MIT]

Some of the most important things in life feel ordinary, until you stop and think about what makes them possible. A cure. A car. A comet. Someone right now is chasing answers to questions we haven't even asked yet. This is #CuriosityOnAMission. 🔬

What are you curious about? https://t.co/rErxJSaMwa

[Sai Nemani @SaiNemani1]

ngl, even though theoretically Gemini 3.5 Flash is smaller than 3.1 Pro, it feels much better to use and a lot less lazy and actually completes the work in detail.

[🚨 AI News | TestingCatalog @testingcatalog]

Google is working on a new Artifact type for NotebookLM called "Lit review". In this mode, NotebookLM will be able to "Generate a Literature Review Matrix" based on your sources.

Considering upcoming additions of Google Play Books and Text Books as sources, Google is planning to push new use cases for readers and writers.

Will it be able to map out all characters from "A Song of Ice and Fire"?

[David @DavidSHolz]

Aristotle said "we should not think human things because we are human, nor mortal things because we are mortal" we should strive to live as though we are more - to "immortalize" and live according to the strongest, best, most supreme elements of ourselves

[OpenAI @OpenAI]

Introducing LifeSciBench, a benchmark for measuring and improving how well AI supports real-world life science research.

Developed with 173 scientists from biotechnology and pharmaceutical research, LifeSciBench includes 750 expert-authored tasks across seven biological research workflows.

https://t.co/JTk0wXHFrT

[antirez @antirez]

Happy to see people reporting GLM 5.2 doing great, the problem is: where to run it, locally? We learned that DeepSeek v4 Flash can lose 50% of the bits and still perform well. PRO seems to also work, but I'm not able to test as much as I could as I like continuous access to an M3 Ultra (but I asked for more continuous access) but if Flash is a proxy, maybe it will work great but needs 512GB of RAM. GLM 5.2 is ~2x the raw weights bits of DeepSeek v4 PRO. Will it ever survive losing 75% of the weights bits without hard damage? I have a hard time believing this will be possible. So great to rent on the cloud, but the current combination of hardware and model size is likely unpractical. Yet: I'll try.

[OpenAI @OpenAI]

GPT-5.4 helped drive a medicinal chemistry project from literature review to a validated experimental result.

Paired with https://t.co/gcDaph8b2B’s Maria AI and specialized lab, the model proposed an unexpected way to improve a widely used reaction in drug discovery.

[Lepore @chatgpt27]

@teortaxesTex trying several times, Expert always tells me May 2026, I don't know, Instant changes the knowledge cutoff every time I ask hahaha

[antirez @antirez]

OpenAI may delay GPT6 (or even 5.6) before making sure could not be blocked like Fable. Or they could play it smart, publishing only the benchmarks that show the improvements on certain area, providing a very censored model in the cyber-security side, and cross their fingers.

[Kimi.ai @Kimi_Moonshot]

🌘 Meet Kimi K2.7 Code HighSpeed!
A high-speed mode of our latest open-source multimodal coding model, Kimi K2.7 Code.

⚡️ Up to 6× faster: Around 180 tok/s on coding tasks with median-length inputs, and up to 260 tok/s on shorter-context tasks.

🔷 Rolling out to Kimi Code Beta Program members, Kimi API developers, and Kimi Business users. (Access will remain limited for now due to capacity constraints.)

🔷 No invite needed. Anyone who joins the Beta Program has a chance to get access 👉 https://t.co/eKogsFGJt6

Open intelligence should be instant, affordable, and borderless. We'll continue improving the model and expanding access as more capacity becomes available!

🔗 Kimi Code: https://t.co/uvoSJKyGCY
🔗 API: https://t.co/mzWxjgGO1h

[Stephanie Chan @scychan_brains]

Really appreciated the chance to give this talk at Harvard Kempner on "What happens after we solve continual learning?"

Slides here: https://t.co/m5xVhSKJBD

[Stephanie Chan @scychan_brains]

"From AGI to ASI": new paper from our team.

This report investigates how AI might develop beyond AGI. It describes theoretical limits, potential pathways, and potential bottlenecks.

https://t.co/x0ZEV2xhNw

[MiniMax (official) @MiniMax_AI]

Run M3 locally today with @UnslothAI

[Unsloth AI @UnslothAI]

MiniMax M3 can now be run locally!🔥

MiniMax-M3 is a new 428B (23B active) open model with 1M context that performs on par with Gemini 3.1 Pro.

Run Dynamic 2-bit GGUF on 138GB RAM/VRAM or 3-bit on 165GB.

GGUF: https://t.co/lwfWsOBNKl
Guide: https://t.co/EP62nmKK0R

[RyanLee @RyanLeeMiniMax]

We’ve been discussing what parameter size works best for the community. While the M3 series boasts a larger parameter count compared to the M2 lineup, we’ve kept its scale deliberately restrained so local model enthusiasts can run it affordably. This time we settled on 428B, hoping it will be accessible to a wider audience.

[Skyler Miao @SkylerMiao7]

MiniMax M3 weights are live.
#1 open-weights model on the AA Intelligence Index.