Feed aggregator

Article URL: https://www.theregister.com/ai-and-ml/2026/08/10/zuck-rekindles-open-weights-llama-drama-with-muse-glimmer/5285666

Comments URL: https://news.ycombinator.com/item?id=49250339

Points: 1

# Comments: 0

Article URL: https://www.youtube.com/watch?v=ffGqjOBowIc

Comments URL: https://news.ycombinator.com/item?id=49250316

Points: 1

# Comments: 0

For three years I worked as CTO of a startup building a huge Postgres app (a WMS) from scratch. We had 100+ inter-related tables and the complexity of delivering feature after feature on this substrate nearly killed us (both the company and our sanity!)

After leaving, I built Typegres to fix the pain with a simpler approach I couldn't find elsewhere:

1. Define your schema: define your schema as TS classes

2. Express your data-model: add methods on those classes (encapsulation 101 applied to your database); this includes both computed columns and relations

3. Expose your API: explicitly mark members to expose

4. Use it (and this is the crazy part): Now the client (e.g., a web browser) can use your data-model directly to query data with SQL-level power (compiling into a single query) over RPC (securely!)

For a visual explanation and more details: https://typegres.com

New features in v0.3:

- SQLite support: complete codebase redo to support multiple dialects with high fidelity

- Live queries: subscribe to complex queries and get fresh results when underlying data changes

GitHub: https://github.com/ryanrasti/typegres Playground: https://typegres.com/play/

Comments URL: https://news.ycombinator.com/item?id=49250266

Points: 1

# Comments: 0

Article URL: https://github.com/tanishqkancharla/calldiff

Comments URL: https://news.ycombinator.com/item?id=49250231

Points: 1

# Comments: 0

Article URL: https://www.nytimes.com/2026/08/10/business/trump-truth-social-traffic-declines.html

Comments URL: https://news.ycombinator.com/item?id=49250230

Points: 3

# Comments: 2

Hello community!

I am here to share a coding agent I have built solo from scratch using agentic engineering. Written in Go, it's a proper coding agent, has features you expect from a useful agent for your daily work, with a minimal and simple UI.

I have named it Keen Code. The repo is here: https://github.com/mochow13/keen-code

Even though it started as an experiment, it is now a full-fledged coding agent for real software engineering work. It supports multiple providers, skills, MCPs, multi-agent orchestration through subagents, automatic compaction, etc.

I have been using it for real production-grade work myself, and also for developing itself.

Notably, I have worked on two separate ideas in this coding agent:

1. Turn Memory

In a multi-turn conversation, tool outputs are removed, only tool call traces are retained. Within a single agent loop, agent sees full tool results but in the next turn, it doesn't see the tool results anymore.

The greatest benefit of this approach is that a lot of tool results that are not needed in following turns don't occupy the context. As a result, context window in a multi-turn conversation with Keen Code fills up much slowly compared to other agents. This is why you will regularly see context window coming down from 20% to 1% at the beginning of a new agent turn.

Of course, this approach has its pros and cons. If agent requires tool result from a previous call, it doesn't have them. But my idea is that tool calls like read, bash, web_fetch are cheap. Do you need to refer to some earlier file you read? Read again. In fact, Claude Code or Codex frequently re-read a file, even though it has read the same file before.

I have plans for more effort in this area. I have a few additional ideas to explore and possibly optimise this approach further.

If you want to read about it: https://mochow13.github.io/keen-code/docs/turn-memory.html

2. Skill-Driven MCP

Another idea I have implemented is skill-driven MCPs. The goal is similar to what Anthropic did with tool-search-tool: optimise context.

In this idea, each MCP server receives a skill. But this skill is not typical "guidance" skill for MCP server usage, rather generated by Keen. Details here: https://mochow13.github.io/keen-code/docs/mcp-skills.html

The big advantage is that no server is pre-loaded completely with tool schemas by default. The agent only receives skill frontmatter for the server. If a server is needed, agent loads the full skill file which lists the tools with descriptions. Then agent reads the specific schema file for the particular tool it wants to invoke.

The drawback is that each MCP call requires file read operation. But everything is locally saved upon discovery, so it's totally fine.

---

Apart from the above two, I am exploring and playing with other well-known context optimisation ideas like hashline edits.

If the above ideas interest you, please do check it out! Here is the CLI usage guideline: https://mochow13.github.io/keen-code/docs/cli-usage.html

Since the project is open-source, issues and contributions are more than welcome!

Comments URL: https://news.ycombinator.com/item?id=49250229

Points: 3

# Comments: 1

Article URL: https://arslan.io/2026/08/10/joining-cursor/

Comments URL: https://news.ycombinator.com/item?id=49250211

Points: 3

# Comments: 0

Article URL: https://reflex.dev/docs/xy/

Comments URL: https://news.ycombinator.com/item?id=49250203

Points: 1

# Comments: 1

Article URL: https://www.bloomberg.com/news/articles/2026-08-10/what-is-sorority-rush-and-why-are-parents-paying-6-500-for-coaches

Comments URL: https://news.ycombinator.com/item?id=49250200

Points: 2

# Comments: 0

Article URL: https://bananarr.com/

Comments URL: https://news.ycombinator.com/item?id=49250176

Points: 1

# Comments: 0

Commented on yesterday's Ask HN: What are you working on? (August 2026), thought I might as well submit a Show HN.

If you see climbing world cup/championship, there is this 3D modelling thing [1]. I want that for my weekly bouldering.

I scanned my local climbing gym into 3D mesh using iphone. Built a simple editor to trim, merge, move/rotate meshes. You can interact with the mesh, view routes, view climb, etc. So this is unlike many gaussian splatting projects where the main use is for viewing. I built a pipeline in updating climbing walls. Climbing routes are manually annotated. Climbing videos are registered against the 3D mesh. Based on one input video, body positions resolved into 3D/4D space. you can view the body landmarks from different angle.

COLMAP, OpenMVS, fastapi, svelte, database is just local json files. A read only build is exported so I can host on github page

If you climb, or know someone who climbs, would love to hear your feedback. Do you usually record your climbs? If I make this a service, what will make you use it?

[1] https://www.youtube.com/shorts/8zdUOaCr6DY

Comments URL: https://news.ycombinator.com/item?id=49250141

Points: 2

# Comments: 0

Article URL: https://blog.helsing.ai/posts/on-comments/

Comments URL: https://news.ycombinator.com/item?id=49250132

Points: 1

# Comments: 0

Article URL: https://artificialanalysis.ai/models/muse-glimmer

Comments URL: https://news.ycombinator.com/item?id=49250123

Points: 1

# Comments: 0

Article URL: https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content

Comments URL: https://news.ycombinator.com/item?id=49250109

Points: 12

# Comments: 5

Article URL: https://moalquraishi.wordpress.com/2026/08/10/a-workaphiles-apology/

Comments URL: https://news.ycombinator.com/item?id=49250104

Points: 1

# Comments: 0

Article URL: https://www.newstribune.com/news/2026/aug/07/toxic-underground-fires-spreading-at-california/

Comments URL: https://news.ycombinator.com/item?id=49250090

Points: 2

# Comments: 0

Article URL: https://www.fastcompany.com/91585494/this-font-looks-perfectly-normal-to-humans-but-confuses-ai

Comments URL: https://news.ycombinator.com/item?id=49249687

Points: 1

# Comments: 0

Article URL: https://allcloud.io/blog/the-great-ai-illusion-why-your-demo-works-but-your-enterprise-agent-fails/

Comments URL: https://news.ycombinator.com/item?id=49249686

Points: 1

# Comments: 0

Article URL: https://bevy.org/news/bevys-sixth-birthday/

Comments URL: https://news.ycombinator.com/item?id=49249681

Points: 1

# Comments: 0