Very cool from a market intelligence perspective, but somewhat concerning from a 'creating adverse incentives' perspective -- e.g., what stops someone from placing bets on these markets then interfering with flights (calling out sick randomly, etc.)?
I feel like it's gotten marginally better (certainly, relative to a few months ago in the subreddits I frequent), but it's unclear how much of that is Reddit, Inc.-driven vs. per-subreddit-moderator-team-driven.
I think the situation can both be "bad" and "better than it was before," but to your point, some of the recent actions (e.g., curating histories) definitely seem counterproductive.
I think the “tool search” and “code mode” counter-arguments against the “MCP eats your context” argument is a bit hand-wavy.
Progressive disclosure has obvious benefits, but I think it would be more valuable to address how agents using either approach can still know which MCP server to search through. Maybe each tool has a search interface, and agents kick off searches in parallel for any ‘yet to be discovered’ tools, but there may be downsides to that as well
Perhaps I'm misunderstanding something, but this article appears to be misleading.
The title and content suggest that it's about AI-native distribution, and that MCP servers are the answer (i.e., the old world prioritized good docs; now, if you want to be discoverable by agents, you need an MCP server). But it assumes agents can just "discover" arbitrary MCP servers (as if Claude would "just know" that you have an MCP server).. which is not the case?
One recommendation would be to create a skill (name_of_your_internal_tool/SKILL.md) describing how the agent should use the tool(s) you're working with. This allows you to progressively disclose that context to the agent rather than filling its context (via its system prompt) with those instructions on every turn. If you use Claude Code, you can also try using it to interactively create the skill (it has a `skill-creator` skill). On the specific skills to make, you could align it with the specific tool (e.g., "Read this skill whenever the user asks you to use X tool...") or with a particular action (e.g., "Read this skill whenever the user asks you to [write, create, etc.]..."
Re: open-source harnesses, ForgeCode appears to be pretty good (currently, #1 on Terminal Bench 2.0 -- https://www.tbench.ai/leaderboard/terminal-bench/2.0). Re: open models, Kimi K2.6 might be a good place to start, but admittedly I'm not too sure how it'd compare to Sonnet / Opus for your use case.
Very cool! Seems like this would be a great way to support group planning (e.g., "work with friend-1-agent, ..., friend-n-agent to [plan a trip, organize a dinner, etc.]")
I didn't notice any signs of AI writing until seeing this comment and re-reading (though I did notice it on the second pass).
That said, I think this article demonstrates that focusing on whether or not an article used AI might be focusing on the wrong “problem.” I appreciate being sensitive to the "smell" (the number of low-effort, AI posts flying around these days has made me sensitive too), but personally, I found this article both (1) easy to read and (2) insightful. I think the number of AI-written content lacking (2) is the problem.
Agreed -- coding agents / LLMs are definitely imperfect, but it's always hard to contextualize "it failed at X" without knowing exactly what X was (or how the agent was instructed to perform X)
Good point – we’ve definitely noticed a lot more Cloudflare representation these days. That said, there seems to be tiers in terms of the protection they offer (and thus the protection used by the websites in this long-tail), where lower tiers (so far) haven’t required proxies.
Curious if you’ve noticed any particularly well defined, obscure websites? Would love to take a look if so.
Haha I appreciate that! And that’s exactly right. Our goal is to make it so that you don’t have to ask the question “but is it worth the time and effort…” when you want to use or explore a new dataset.