Hacker News - Newest: """model" "context" "protocol"" "MCP""

Show HN: Rafter – an MCP server that shares one team's memory, skills and agents

kristohb — Fri, 28 Aug 2026 08:45:35 +0000

Supercharge your team with Rafter at https://heyrafter.xyz! :rocket:

Your team members is spending too much time and tokens in their AI tools just to solve a trivial problem. They copy data from drive, jira, onedrive and they have some prompts they paste in. Every time they go back and forth explaining how the problem should be solved. For some this is the reason not to use AI.

It's complicated.

Rafter allows you team to share agents, skills and memory Connect your tools. The knowledge graph gets built allowing your team to stay in flow. Works in ChatGPT, Claude, Gemini, Codex, Claude Code, Cursor, Opencode etc

Go to https://heyrafter.xyz/#get-started to setup Rafter in your favorite AI tool.

Comments URL: https://news.ycombinator.com/item?id=49476048

Points: 1

# Comments: 0

Elon Musk's xAI used child porn to train Grok models, lawsuit says

KingMob — Fri, 28 Aug 2026 08:20:41 +0000

Article URL: https://arstechnica.com/tech-policy/2026/08/elon-musks-xai-used-child-porn-to-train-grok-models-lawsuit-says/

Comments URL: https://news.ycombinator.com/item?id=49475883

Points: 4

# Comments: 2

I made a game with ads you can remove by beating the hard mode once

shadylabs — Fri, 28 Aug 2026 08:19:24 +0000

Article URL: https://apps.apple.com/ie/app/qube-mobile/id6797254645

Comments URL: https://news.ycombinator.com/item?id=49475873

Points: 1

# Comments: 0

Agent Communication Context Protocol (ACCP)- Proposal

amonte — Fri, 28 Aug 2026 07:26:28 +0000

Article URL: https://github.com/RupertCloud/Agentmail/blob/main/docs/accp/SPEC.md

Comments URL: https://news.ycombinator.com/item?id=49475461

Points: 1

# Comments: 0

Fuzzy terminal manager to pick, configure, and run local GGUF models

modinfo — Fri, 28 Aug 2026 07:01:39 +0000

Article URL: https://github.com/skorotkiewicz/run-model

Comments URL: https://news.ycombinator.com/item?id=49475286

Points: 1

# Comments: 0

When You Need an MCP Server and When You Don't

jasondoyle — Fri, 28 Aug 2026 05:43:08 +0000

Article URL: https://jasondoyle.ie/whitepapers/when-you-need-an-mcp-server/

Comments URL: https://news.ycombinator.com/item?id=49474788

Points: 2

# Comments: 0

Entropy – Mental Model

fagnerbrack — Fri, 28 Aug 2026 04:01:30 +0000

Article URL: https://read.perspectiveship.com/p/entropy

Comments URL: https://news.ycombinator.com/item?id=49474209

Points: 1

# Comments: 0

A second model pays off in two shapes

Marvin_RunAI — Fri, 28 Aug 2026 01:49:28 +0000

The claim is Anthropic's, from their cost-optimization guide: a second model paid off in two shapes, an advisor and an orchestrator. Nothing else they measured made the list.

Comments URL: https://news.ycombinator.com/item?id=49473442

Points: 1

# Comments: 0

NexaQL A binary protocol that bridges legacy SOAP/XML with modern Edge Computing

ysaias2028 — Thu, 27 Aug 2026 23:51:14 +0000

Article URL: https://github.com/YsaiasPeru/nexuql

Comments URL: https://news.ycombinator.com/item?id=49472717

Points: 3

# Comments: 1

Ask HN: Which class of business model is most resilient to build in the AI era?

chatmasta — Thu, 27 Aug 2026 23:32:33 +0000

If you were to start a company today, what type of business model would you choose? Most new businesses have some form of AI in them, and this makes some sense during a gold rush, but it’s also a crapshoot. And the rate of change is so insanely fast and unpredictable that it’s becoming difficult to read the market with any confidence beyond three months. Even if you hit traction, there’s no guarantee your moat won’t evaporate. And it’s cheap for everyone to build now.

Personally I think the smartest business to start right now would be an “online-to-offline” model that bridges digital and physical world - think DoorDash, Uber, etc… if you hit traction with that kind of business then at least you have a legitimate moat that AI can’t drain overnight.

Comments URL: https://news.ycombinator.com/item?id=49472620

Points: 3

# Comments: 5

What the heck is a context layer

Debanitrkl — Thu, 27 Aug 2026 22:18:47 +0000

Article URL: https://nexla.com/what-is-a-context-layer/

Comments URL: https://news.ycombinator.com/item?id=49472035

Points: 1

# Comments: 0

Running local large language models: not as difficult as you might think

sebg — Thu, 27 Aug 2026 22:13:43 +0000

Article URL: https://vitaexmachina.substack.com/p/running-local-large-language-models

Comments URL: https://news.ycombinator.com/item?id=49471992

Points: 1

# Comments: 0

Show HN: IQ Routing – route every call to the cheapest model that holds quality

Georgeavvila — Thu, 27 Aug 2026 21:40:02 +0000

Article URL: https://iq-routing.com

Comments URL: https://news.ycombinator.com/item?id=49471637

Points: 1

# Comments: 0

Model Success Rate May Count the Wrong Runs

DISCURSIVE — Thu, 27 Aug 2026 21:29:58 +0000

Article URL: https://topicqueue.substack.com/p/four-new-benchmarks-say-your-success

Comments URL: https://news.ycombinator.com/item?id=49471527

Points: 1

# Comments: 0

Show HN: We built open OpenRouter that turns usage into a better model

SilenN — Thu, 27 Aug 2026 21:18:35 +0000

Hi HN, we built an open source model gateway. It's a single place to manage our own self hosted, frontier, and open source models in one place.

It’s is rust native, built for concurrency, and implements all the config quirks across models and providers (streaming formats, tool calls, model parameters, rate limits, and different error behavior).

The gateway adds under 1 ms for BYOK requests and under 2 ms when Experiential supplies the provider key. It has every major inference provider, and 1000+ models refreshed daily via a codex agent that opens a PR.

Compared to other similar projects we’re open source, take no markup, allow you to mix local models with a marketplace, and use your traffic to (opt in) train you a model. Simple routing doesn’t warrant a 10% token markup.

The way we do this is given standardized OTel traces, we mine representative real tasks, use text world models to simulate rollouts for various models, apply an LLM judge, and fit a nearest neighbor classifier on top of an embedding of a prompt to decide the optimal model for each request. Usually this can map out a better pareto curve on cost/quality than just calling single models but it’s not perfect.

Using these simulations we can also do things like suggesting cache hit optimizations, new model suggestions, and training models.

It’s open source, so you can deploy it on your own infrastructure, use our hosted version with 0 markup, or read how we design for maximum availability on our website.

Comments URL: https://news.ycombinator.com/item?id=49471407

Points: 175

# Comments: 33

Unlocking the Power of AI in Visual Studio with Bring Your Own Model

ibobev — Thu, 27 Aug 2026 21:04:32 +0000

Article URL: https://devblogs.microsoft.com/visualstudio/unlocking-the-power-of-ai-for-every-developer-in-visual-studio-with-bring-your-own-model/

Comments URL: https://news.ycombinator.com/item?id=49471240

Points: 2

# Comments: 0

Anthropic Announces Model Hardware Standard, MCP for Hardware?

luigipederzani — Thu, 27 Aug 2026 21:02:12 +0000

Article URL: https://modelhardwarestandard.com/

Comments URL: https://news.ycombinator.com/item?id=49471216

Points: 1

# Comments: 1

Show HN: Context-Aware Header Audit; Built for Learning, Not for Scores (German)

Hacker News - Newest: """model" "context" "protocol"" "MCP""

Show HN: Rafter – an MCP server that shares one team's memory, skills and agents

Elon Musk's xAI used child porn to train Grok models, lawsuit says

I made a game with ads you can remove by beating the hard mode once

Agent Communication Context Protocol (ACCP)- Proposal

Fuzzy terminal manager to pick, configure, and run local GGUF models

When You Need an MCP Server and When You Don't

Entropy – Mental Model

A second model pays off in two shapes

NexaQL A binary protocol that bridges legacy SOAP/XML with modern Edge Computing

Ask HN: Which class of business model is most resilient to build in the AI era?

What the heck is a context layer

Running local large language models: not as difficult as you might think

Show HN: IQ Routing – route every call to the cheapest model that holds quality

Model Success Rate May Count the Wrong Runs

Show HN: We built open OpenRouter that turns usage into a better model

Unlocking the Power of AI in Visual Studio with Bring Your Own Model

Anthropic Announces Model Hardware Standard, MCP for Hardware?

Show HN: Context-Aware Header Audit; Built for Learning, Not for Scores (German)

Sopro V2: SOTA voice cloning TTS model that runs on your CPU

Show HN: Discover and share AI prompts and MCP servers