ngage

@hexednobility @miketuritzin yeah that’s definitely wrong. Assuming you have a “clear system” upfront is always a great way to undermodel your assumptions

This is an underrated way AI will change software. It’s now silly to not implement the general system up front because once you have a clear system with parameters and appropriate abstractions its much easier/safer for LLM to augment in the future. Plus its much quicker to implement in the first place.

A programming lesson that's been hammered into my skull is to not try to create a general system until you've implemented AT LEAST a handful of one-offs the system is supposed to support. Otherwise you waste endless time fighting with and refactoring your "general" system.

@VictorTaelin @francidellamora @alexalbert__ vouch

@alexalbert__ I'm the maintainer of Bend, a new programming language with 19k+ stars on GitHub. We're about to launch a major update. Having access to this model to audit it would greatly improve the project's security, and of projects built with it. Lmk if there's any way to get involved.

We released Claude Opus 4.6 just two months ago. Today we're sharing some info on our new model, Claude Mythos Preview.

new personality just dropped 🤝🌊

X notifications are unbelievably broken @nikitabier

why can some guy from bumfuck nowhere notify me with this garbage nuclear war bait @nikitabier

@gpusteve 17 had some bangers

realized there was genuinely levels to ts when a quant firm asked me my favorite version of c++

@imalicenine Sym S tier or riot

the most important overwatch video of all time just dropped

@Nehmoe_ @Tricendo_ Yeah, likewise

@nathan___gage @Tricendo_ Windows update im on idk the number but im on the one from January or December and ive been waiting to install the others cause Microsoft keeps having issues that crash people's pcs. Nvidia is latest. Ive rolled back both windows and nvidia updates though to check. Refresh 480hz

This patch my fps in vaxta stays between 240-300. I used to be able to get 600 in vaxta with no issues. My GPU, CPU, and memory usage are not even close to bottlenecking. from all low settings with 50% render scale, to 200% nothing changes. Overwatch PLEASE fix the performance

@Tricendo_ @Nehmoe_ Thanks man

@nathan___gage @Nehmoe_ 591.86 nvd, win11 25H2 os build 26200.8037

@Tricendo_ @Nehmoe_ Do you know what nvidia driver you’re on, as well as windows update + refresh rate?

@Nehmoe_ i have a 5070ti, 5800x3d and i avg 587fps at 1440p and 1 % lows of 468

@burkov @ParkerSligting @bcherny the entire file format being Microsoft’s docx is the branding

@ParkerSligting @bcherny Imagine Microsoft integrating "Mistypings fixed by Microsoft Word" into all Word documents!

When Anthropic adds "Built with Claude" to every commit, it looks as dumb as when iPhone users send their emails with the "Sent from my iPhone" signature.

Humans don’t even write prompts anymore they just look at the terminal like this

@Richytoons @dafran 6v6 has a 2 tank maximum restriction; queue times are great with a low player base as-is.

@dafran Are we talking before role lock or after? Because I agree, but with no role lock 6v6 we end up with Goats or worse and with 30 min queues. I think what they mean by 5v5 is a more controlled environment and more fun in general

The biggest reason 5v5 is more popular, is because its the main game mode. I am not going to play 6v6, when all the pro players play 5v5. If 6v6 was the main game mode and had role q and the pro players played 6v6, it would be far more popular than 5v5 and 5v5 would die out

@kianbazza drop 'area'. just 'hit'. its cleaner.

You did it. I'm second-guessing myself now... Should 𝚑𝚒𝚝-𝚊𝚛𝚎𝚊 be called something else? I wonder if 𝚑𝚒𝚝𝚋𝚘𝚡 is better 🤔 Or even shorter, 𝚑𝚋? For reference: 𝚑𝚒𝚝-𝚊𝚛𝚎𝚊-𝟷 𝚑𝚒𝚝-𝚊𝚛𝚎𝚊-𝚢-𝟸 𝚑𝚒𝚝-𝚊𝚛𝚎𝚊-𝚕-𝟹 𝚑𝚒𝚝𝚋𝚘𝚡-𝟷 𝚑𝚒𝚝𝚋𝚘𝚡-𝚢-𝟸 𝚑𝚒𝚝𝚋𝚘𝚡-𝚕-𝟹 𝚑𝚋-𝟷 𝚑𝚋-𝚢-𝟸 𝚑𝚋-𝚕-𝟹 Let me know what you think. ⤵️

@jeffzwang PTC / codemode lighten the load w.r.t. context engineering on server-side. MCP clients just haven't widely adopted the structured output return signatures, yet.

In my opinion, MCP is not redundant and bad performance owes mainly to skill issue: Redundancy The most important leverage an agent can have is amazing context engineering, and CLIs/APIs simply aren't designed to expose precise, concise interfaces for specific use cases. To achieve that, you need a client-server relationship where the client (agent) tells a server about itself and the server exposes an intelligent selection of capabilities (tools), or if the client (agent) post-filters the set of tools (in MCP, what is returned by `initialize`). Well, MCP is such a technology! Could you hack CLIs/APIs to achieve the same thing? Sure, but they're not designed for selective exposure that minimizes an agent's context pollution. You need some new protocol layer thing that allows for that, even if that just wraps CLIs/APIs, because of this agent-specific problem. MCP! Performance Most of the performance issues I see in MCPs owe to bad implementation. That they promote good context engineering is a moot point if they're designed such that context gets mega polluted, and unfortunately many MCPs mega pollute context. One way this happens is exposing way too many tools - it's pretty common right now for API-wrapping MCPs to just expose every single API endpoint and request field, with some MCPs in the tens of thousands of tokens size because of this. This wouldn't be an issue if agent harnesses allowed for selective tool filtering per MCP, but most agent harnesses (e.g., Claude Code) don't support this. Also developers are not mindful enough of tool call response token payloads, which also need to be minimized to avoid context pollution - it is all too common for responses to constitute of the full API response payload. This has been partially mitigated by the rise of subagents but is still a major issue that often APIs/CLI based tool implementations may not have because in those implementations there are existing patterns to filter payloads. In my personal Claude Code setup, I address these problems by 1) implementing an "MCP proxy" layer that lets me manually remove tools from MCPs, 2) implementing post tool call hooks that filter response payloads. And in the Exa MCP, we are careful to 1) return only fields that an agent needs 2) filter parsed HTML to only relevant tokens using models that we train. But these features need to be built into MCPS and agent harnesses as first class citizens. Lastly, I will note that it is obviously the case that agents know how to use APIs/CLIs that are in their training data very well, almost always better than MCPs. But that issue should get fixed over time as LLMs get trained on more MCP data. Overall I tend to be relatively pro when it comes to new shit that needs to be built specifically for AIs (after all, our company is in the business of building a search engine from scratch for AIs lol). But this is my experience as someone who has thought about this way too much. Curious for people's thoughts! Maybe I am missing something.

if your startup's chat didn't look like this at exactly 9:46:17pm EDT then you're ngmi

somehow claude code created the best memory feature ive used - im a real memory hater but they did it quite tastefully

My weirdest belief is that science demystifies the world and strips it of divinity until you spontaneously achieve gnosis and realize that material reality IS inextricably divine

The proposition frames atheistic rationalism not as tabula rasa neutrality but as a phylogenetic offshoot of Christian metaphysics: a secularized soteriology wherein salvific telos migrates from divine grace to human instrumental reason, preserving eschatological linearity (progress via scientism) while excising the transcendent anchor. This inversion engenders epistemic hubris—the rationalist ego posits itself as Laplacean demon, capable of exhaustive Bayesian world-modeling sans embedded cultural heuristics, tacit knowledge reservoirs, or distributed cognition (cf. Polanyi’s personal knowledge or Oakeshott’s practical rationality). Redditor ontology thus recapitulates the very fideistic dogmatism it abjures, mistaking its inherited post-Protestant individualism for primordial default while pathologizing tradition as mere “bias.” The meta-narrative of “enlightened default” functions as cultic apologia, immunizing against self-reflexive critique of its materialist priors.