Where the AI market is talking

Google is making private AI practical with homomorphic encryption

What does that have to do with incpa? I send this server a question encrypted. It stays encrypted yet the server distinguished my ciphertext enough to produce a coherent reply. That means there was enough information in the ciphertext for the server to perform the inference, even if it was obscured to where it looks indistinguishable, it apparently wasn't. This doesn't remove trust from the equation, it puts the trust directly on the algorithm. Probably one no one can explain to me like I'm 5. But because it's encrypted I am no longer breaking the law by processing on protected data. I imagine very large contracts await on the other side. Lots of money to be made.

Why does Opus 5 feel worse to work with?

I was the same until I ran out of Anthropic tokens one day and used "Grok Build" which is their Claude Code clone. You can use config to point it any LLM API so don't need to use Grok, and I like the UI better too.

We urgently need a coherent national AI cybersecurity policy

The US urgently needs coherent national policy; add AI to the (long) list

Show HN: Captain, AI Travel Agent

I’m a software designer from Lagos exploring conversational interfaces and frameworks for building agents. I made Captain to help with everyday travel planning. It can explore an itinerary, check live flight prices, or watch for the right time to book. Captain is a Telegram bot powered by a generalist model (anthropic/claude-sonnet-5), with specialist models handling workflows such as trip interpretation and voice transcription. It also includes a visual workspace where users can review the agent’s work and directly manipulate their travel information. I'm building the agent to explore how conversation, tools, durable workflows, and direct manipulation can come together in one coherent product.

Show HN: Captain, AI Travel Agent

I’m a software designer from Lagos exploring conversational interfaces and frameworks for building agents. I made Captain to help with everyday travel planning. It can explore an itinerary, check live flight prices, or watch for the right time to book. Captain is a Telegram bot powered by a generalist model (anthropic/claude-sonnet-5), with specialist models handling workflows such as trip interpretation and voice transcription. It also includes a visual workspace where users can review the agent’s work and directly manipulate their travel information. I'm building the agent to explore how conversation, tools, durable workflows, and direct manipulation can come together in one coherent product.

Maximizing the value of your Claude Code sessions

This feels like the Anthropic version of "You're holding it wrong" (1) 1) https://www.wired.com/2010/06/iphone-4-holding-it-wrong/

Qwen 3.8 27B

Trading blows in some benchmarks is a bit exaggerated. If you try the model, `xhigh` is basically feels like the `max` mode (i.e. massive thinker and extremely presistent), and the amount of world knowledge and intent understanding is nowhere close to an Opus class model even from 6mo ago. It's still very useful, and it'll probably displace a good bit of API spend; but it's not really "trading blows with SOTA from just a half year ago". A bit overblown on the Anthropic/OpenAI has missed the window I reckon. Also, on the open weight frontier side, Kimi K3 is pretty expensive, and Deepseek V4 Pro/Flash is getting a little less juicy with price increases.

We urgently need a coherent national AI cybersecurity policy

Show HN: APIMart: Discounted AI API Aggregator for GPT-5, Sora 2

I'm pretty sure enterprise customers can also negotiate lower rates But OpenAI and Anthropic obviously wouldn't give better rates to a reseller. If the service is operating legally (no stolen accounts or the like, just breaking ToS) it is almost certainly proxying coding plans

Qwen 3.8 27B

I wonder if Anthropic and OpenAI possibly missed the window to go public. A 27B open-weight model trading blows with the SOTA from just half a year ago is not great news for trillion-dollar investments...

Why does Opus 5 feel worse to work with?

please god I want anthropic to fail, so that they could learn that their current approach is wrong

Why does Opus 5 feel worse to work with?

The entire Hugging Face hack involved escaping major sandboxes, this emergent (or intended) behavior in a smaller scale is still a real issue.

Qwen 3.8 27B

Yeah, I might just replicate what you're doing. Main issue right now is just finding spare vram to actually run another model in parallel... And yeah, if I train up a vision adapter somehow, I'll try to put it up/post about it, seems like we're getting the killer apps for local LLMs right now, where it's just feasible enough if you're enthusiastic enough to be a bit economically irrational, and just useful enough to sort of rationalize.

Training AI Scientists to Replicate Research

Ask HN: How much money do you spend monthly on subscriptions for AI models?

Just $20 a month to Anthropic. I've run into having to step away from the AI projects for a bit here and there due to burning the 5-hour window in 3 hours a handful of times. But never ran out of the weekly usage with days to go.

Fortunes from AI revive effective altruism after Sam Bankman-Fried turmoil

Archive: https://archive.is/20260814045441/https://www.ft.com/content... The EA folks are at it again > The FT has identified more than 60 current and former Anthropic employees who have pledged on the “Giving What We Can” platform to donate at least 10 per cent of their income, reflecting effective altruism’s emphasis on using evidence and data to maximise the impact of philanthropy. Anthropic’s seven co-founders, who each have an estimated worth of about $8bn, have separately pledged to give away 80 per cent of their wealth, creating the prospect of billions of dollars in future donations. > The prospect of new AI-created wealth is breathing new life into a movement that was forced into a reckoning after FTX’s former chief Bankman-Fried — one of …

Qwen 3.8 27B

Similar to the way they asked Sol to solve Erdos problems, that's what I want my model to do for programming. I don't want to try to take my best educated guess at what the best design is BEFORE implementation - especially if you're designing a feature for a codebase you're not an expert in, you don't know like the back of your hand (i.e. one that is mostly or entirely LLM generated). What sounds good on paper - often times becomes unideal in practice when you get to the reality of implementation. It may not be worth re-architecting your entire system to get to a "pure" design that would be the best - all things considered. Instead, I'd like the model to independently design many plausible and coherent good solutions, then implement each of them, then intelligently pick the few winners (a…

Why does Opus 5 feel worse to work with?

> according to Anthropic and Google, it doesn’t really impact the quality of responses. I find that a bit hard to believe, although those guys are much smarter than I. I would guess "it doesn't impact the quality of responses" was guaranteed to be claimed before they even implemented any of the watermarking. And would come from marketing, not the people who implemented it.

The Qdrant Output Connector

When Genius Fails: The Intellectual Arrogance of the AI Labs

This is only true if newer investors get pro-rata ownership of the Anthropic stake. If that position is side-pocketed, then any investors who were not there at the time of the Anthropic allocation are looking at around 90% loss on their investment.