Vibe Coding Community | Digg | Digg

2d

1Password open sources a benchmark to stop AI agents from leaking credentials - Help Net Security

1Password has open-sourced a benchmark called SCAM to evaluate AI agents' ability to avoid phishing attacks in real workflows. The benchmark simulates workplace scenarios and tests whether AI agents can recognize and avoid common threats without being prompted. After baseline testing, 1Password introduced a 'security skill' document to improve agent risk assessment, significantly reducing critical failures. The benchmark is available on GitHub under the MIT License.

2d

Not far from truth sometimes

@jeremylongshore

5d

Anyone ?

Finding a Technical Co-Pilot for Sales Calls (Fractional) I’m looking for a senior engineer to join prospect calls as a fractional technical sales partner—someone who can quickly learn the product, translate architecture decisions for CTOs/engineering leads, and help close dea

10d

Have you tried Opus 4.6 yet?

Opus 4.6 is state-of-the-art on several evaluations including agentic coding, multi-discipline reasoning, knowledge work, and agentic search. We're also shipping new features across Claude in Excel, Claude in PowerPoint, Claude Code, and our API to let Opus 4.6 do even more.

11d

Digg Sidekick Build Notes Day 0+1

Project: Digg browser extensionBuild time: ~1:15 hourStatus: Functional prototype, Chrome Web Store pending Repo: https://github.com/timmolter/Digg-Sidekick Build Process Started with a single AI prompt from DeepSeek to Opus that generated a complete, working extension. Keyw

12d

Just launched an open-source Digg extension—built with AI 🤖

Hey /aidevs, Been tinkering with a browser extension for Digg and decided to open it up early. It's a simple filter tool - I'm calling Digg Sidekick for now. 🔗 Repo: https://github.com/timmolter/Digg-Sidekick Built it mostly with AI—I'll write up the prompts and process

12d

I was thinking of making a Digg extension in a "build in public" style in this community

Would anyone be interested in following along? I would host it on a git repo.

13d

Anthropic's Sholto Douglas, foreshadowing that the new Sonnet models end up being smarter than Opus models.

13d

How many do you think you could keep in check in parallel??

OpenAI's new Codex app lets you manage multiple AI coding agents in parallel. They built a full racing game from ONE prompt using 7M tokens. We've officially entered the developer-as-project-manager era.

@fillosofillkall

13d

Can We Just Make "Secure Vibe Coding"?

A big criticism of vibe coding is that it is "insecure"so... can't we just run a bunch of security tests and have LLMs audit code more, for stronger security?Why hasn't there been more attention to this kind of thing, do you think?

14d

Claude Sonnet 5 drops next week.

• 1M context • Half the price of Opus 4.5 • Best-in-class for agentic coding

15d

The 80% Problem in Agentic Coding - by Addy Osmani

The article discusses the shift in software development practices due to the increasing use of AI in coding, highlighting the '80% problem' where AI can generate 80% of the code but leaves the final 20% to humans, which is more complex and requires deep understanding. It emphasizes the importance of maintaining comprehension and quality control as AI adoption increases, and suggests that successful developers will be those who can effectively manage and direct AI tools rather than those who can write the most code. The article also touches on the potential for a 'slopacolypse' where the quality of code may decline due to over-reliance on AI, and the need for robust verification processes and architectural hygiene to mitigate these risks.

16d

On behalf of /build, I'm gifting some community icons

I'm making these icons to be in line with the art style that Digg uses for communities like /digg and /music Here are the unsolicited community icons created specifically for /aidevs! Feel free to use or toss (or I'm happy to revise!). Thanks! Rylet P.S. If you had some

6 images

16d

Looks good to me” — the most dangerous phrase in software

17d

Claude Code Opus 4.5 Performance Tracker | Marginlab

Claude Code Opus 4.5 Performance Tracker The goal of this tracker is to detect statistically significant degradations in Claude Code with Opus 4.5 performance on SWE tasks. •Updated daily: Daily benchmarks on a curated subset of SWE-Bench-Pro •Detect degradation: Statistical t

17d

OpenAI will retire several models, including GPT-4o, from ChatGPT next month

OpenAI announced it will retire several models from its ChatGPT chatbot next month. GPT‑4o, which is beloved for its warm conversational style, is one of the models that will be removed from the chatbot. OpenAI said just 0.1% of users are choosing to use GPT-4o every day.

17d

We are starting to test ads in ChatGPT free and Go (new $8/month option) tiers.

https://x.com/sama/status/2012253252771824074

18d

Was Cline just acqui-hired by OpenAI?

Kilo is celebrating the Cline team's success in being potentially acqui-hired by OpenAI, while emphasizing their commitment to keeping Kilo open-source and supporting Cline contributors and users.

18d

I work with models

18d

Kimi K2.5 is free for a week on Kilo Code. This model beats Opus 4.5 on several coding benchmarks.

Saw this on twitter: https://x.com/kilocode/status/2016449095511007535and tried it. I'll update later after using it for a while. Have you tried it? What do you think?

19d

Secrets: In the /digglist Growth Chart Tool, if you look at the HTML, you can see the ClaudeAI prompts used to make it

Tool: http://digglist.com/search_chart.html I used ClaudeAI to quickly make an HTML chart page. I also made it so that your computer does the work, not an SQL backend. This caches faster for both you and the server.

19d

The challenge of making AI conversation the ultimate "Data Cleaning" tool.

I'm building Dashform, and my goal is to turn the "messy reality" of user input into "clean, structured database rows"—all through a simple chat. Instead of forcing users to format their own data (drop-downs, specific date formats), I want the conversation itself to be the clean

19d

Happy or Sad for the death of stackoverflow??

monthly stack overflow questions over time. 3710 questions last month, just slightly under the 3749 from the first month of it being public. human software engineering had a good run, and now we've come full circle.

20d

OpenAI’s president is a Trump mega-donor

Greg Brockman, president of OpenAI, and his wife donated $25 million to the pro-Trump super PAC 'MAGA Inc.' in 2025, making up nearly one-fourth of the six-month fundraising cycle. This donation is part of a larger trend of tech executives supporting President Trump's administration, which is actively backing the AI industry and opposing state-level regulations. Brockman has also supported the pro-AI super PAC 'Leading the Future,' which has targeted New York State Assemblymember Alex Bores, a cosponsor of the RAISE Act. This political involvement by tech leaders has coincided with the administration's efforts to roll back consumer protections and tech regulation.

20d

New Name for Clawdbot ==> Moltbot

The name change was prompted by trademark concerns from Anthropic, as “Clawdbot” and its mascot “Clawd” were deemed too close to the Claude AI brand. Rather than push back, the project embraced a thoughtful rebrand aligned with its lobster theme: molting symbolizes growth, evolut

20d

Sharing a decent inforgraphic I found about Clawd

... since I was wondering myself exactly whatt he hype was all about.

21d

How far have you taken a fully vibe coded project?

Proof of concept?MVP?In market making money? A functioning side hustle?Curious to know where things worked and where they fell over for you.

21d

The reason why RAM has become so expensive :)

22d

Wikipedia parent partners with Amazon, Meta, Perplexity on AI access

Wikipedia created a direct API for LLMs to pull training data from. Pretty cool IMO!

23d

Hello World

Obligatory hello world post to welcome everyone to this community! 0101001001101000100100100