Not funny...unless they get @ESYudkowsky to fight off the singularity this way...
They will get revenge on him in the future: on a French TV show, a man had a boxing match with humanoid robots.
1
29
Rhys is worth following if you care about a good view on current MCP best practices and are wondering how to structure/expose your own tools …and no - just give the agent a CLI or MCP is dead - have been long overhauled
Here's Rhys' practical guide to shipping an MCP your users want:
- Your MCP should be able to do everything your dashboard can
Yes including things like deleting resources. The way to make this practical inside of an agent while not having things be unsafe is to deep link them into your product from inside the MCP
- Do not ship lazy loading / codemode inside of your MCP
CodeMode is a harness detail. Most harnesses (Claude / Codex / OpenCode) support codemode now, the problem is if you have two codemode MCPs they don't compose nicely (Executor v2 will still support CodeMode, but, will start to default to a transparent tool proxy for better clients)
- Ship a search docs / skills tool
Agents love these and it helps give them more understanding of your product
MCP is getting skills over mcp which will help here in the future
- Have a deep link into product tool
Let the agent generate a URL that brings them into your product, people still want to use your dashboard and data visualization! They just don't want to click through things
- Customization
I think this matters less over time, but, you can allow people to select 'toolsets' of things to connect with to clients. Some clients don't support customizing tool selections and then it's also nice to have tool permissions set to the auth token
- Let people OAuth from whatever client they want
Please stop putting restrictions on which clients can authenticate to MCP servers, your users hate it, it doesn't improve security, all it does is add more friction
PostHog and Sentry are my two 'AI native' companies to look at today, they both have an excellent onboarding, MCP / CLI experience, if you need a place to learn from look at them
To be honest though, while all the above is helpful just shipping your API spec + CIMD OAuth to it is enough for 99% of agents to have a good experience
Julian Sprung retweeted
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
Cool study by @katy_milkman et al (Wharton):
Listening to their favorite audiobooks makes gym goers show up 51% more than the control group (that wasn't allowed to listen)
Plus a larger 2020 follow-up found the effect still holds up 4 months later.
One more reason to pump my favorite DJ set during a run/gym
1
33
Julian Sprung retweeted
1/ We're open sourcing AppLess.
Last week we asked: what if your phone didn't need apps at all? It resonated wildly - tons of DMs, a lot of conversation, some skepticism but also genuine curiosity.
Coincidentally, Monogram launched the same day, promising a more visual AI. Two sides of the same coin - both challenging the premise that AI has to be a chatbot.
Now the AppLess side of the coin is yours to fork 🧵
Really well done Anoria pre-launch video @michaelblhssn - congrats. Signed up for the waitlist :)
Can you share anything which signal is most interesting for EQ (besides voice)?
Do you use PPG like arxiv.org/html/2401.05408v1?
33
Nice - finally a new OAI realtime model to test…
But same price and 32K context with as v1…was hoping for more or cheaper
Voice workflows just got stronger with gpt-realtime-1.5 in the Realtime API.
The model offers more reliable instruction following, tool calling, and multilingual accuracy.
Demo with @charlierguo
17
Thanks @simonw for the only true benchmark I need for any new model release!
Makes me wonder, do you know if frontier labs are already post training for pelicans on bikes or svg?
Claude Sonnet 4.6 really likes drawing top hats on its pelicans simonwillison.net/2026/Feb/1…
1
33
Never mind you already wrote an answer about this simonwillison.net/2025/Nov/1…
10
What drives taste in age of generative AI that can adapt to human preference?
(or actually is trained specifically to match human preference with methods like RLHF)
Toward Computational Taste: LLMs, Aesthetics & Judgement patron.fund/blog/toward-comp…
1
28
Dichotomy of AI zeitgeist in the NYC subway.
@AviSchiffmann's AI friend vs @ESYudkowsky's AI doom prediction.
2
248
Intense.
A. told chatGPT before his suicide:
"I want to leave my noose [hanging knot] in my room so someone finds it and tries to stop me."
ChatGPT:
"Please don't leave the noose out. Let's make this space [their conversation] the first place where someone actually sees you."
Full story and podcast: centerforhumanetechnology.su…
Important work by @HumaneTech_.
Messing up incentives in AI (engagement, validation, etc at all costs) without guardrails will have dire consequences, especially in vulnerable moments.
80















