Opus Clip charges $29/mo to guess. This free open-source tool ranks instead.
Paste a YouTube link, it finds the best moments and scores every clip for virality. Bring your own API key, pay for tokens only.
https://t.co/6GclLQwX86
Gamma doesn't know your client. Claude does.
Claude Slides turns a week of project chat into a 10-slide deck from one prompt. Every element editable, exports to PowerPoint, copies the design of any old deck you drop in.
Rolling out on Pro and Max: https://t.co/Nben0Xb4Ni
This new AI model is 444x cheaper and 193x faster, because it never writes a sentence.
Jev: you give it the options, it picks one in under half a second. It can't make stuff up, the answer has to be on your list.
Hand it the boring jobs. Waitlist:
https://t.co/Jlytqm3kGJ
You can now learn anything with AI.
anything2explainer is a free Claude skill: type a topic, it researches real sources, writes the script, generates the voiceover, animates every shot in code and adds subtitles.
A tutorial made just for you.
https://t.co/FisCAqQpKG
Claude can now run your entire LinkedIn for you. Free.
linkedin-agent-skill: 11 Claude skills. One writes posts from 21 hook formulas, one comments on others' posts, one replies in your voice.
Prompt it once, your week is planned.
https://t.co/E8N1ZZo6SC
🚨 A legal bug-bounty test reportedly took Claude Opus 5 about three hours to reach OpenAI’s private code repo.
This episode breaks down what happened, why a connected employee account could reach private company code, and what it means as models get better at breaching defenses.
Then: Claude Projects as a cloud chief of staff, agent orchestration, Astra for law, AI-assisted research, recursive self-improvement, and the creeping feeling of becoming AI numb.
@nicksaraev × @AIJackRoberts
cover AI security, bug bounties, cloud agent orchestration, context, legal AI, model versus application layers, AI-assisted research, recursive self-improvement, and the psychology of AI acceleration.
0:00 OpenAI bug bounty
5:17 Small bounty and security
6:57 Claude Projects and agents
8:37 Agent orchestration and context
12:33 Astra for law
13:58 Model versus application
16:44 Claude’s supervised research
19:43 Recursive self-improvement
20:59 Mac mini giveaway
23:35 Fruit fly race winner
25:20 Last week’s winner Daniel
27:01 Questions on AI numbness
30:54 Fast classification in AI
32:21 Outro
Nvidia scanned 40,000+ Claude skills. 1 in 20 was malicious, a quarter had security holes.
So they built SkillSpector, free and open source: 71 checks, point it at a GitHub link before you install anything.
https://t.co/ULVPevGBWK
Anthropic just released 11 free plugins that each already know a job: sales, marketing, finance, legal, support.
Not prompts. Each one knows how the role works and is wired into the tools it uses. Install once, stop re-explaining.
https://t.co/XjidLT1tkx
For anyone running both Astra and Fable... I'm leaning Astra day to day and still opening Fable 5.1 for anything that needs real reasoning.
which one do you actually open first, and what made you pick?
I tested Astra vs Fable: I gave both models deliberately terrible prompts. On purpose.
I wanted to see them work under bad conditions. a worse model with the right system beats a better model with a lazy prompt.
Ten months of private Claude Code tuning just went open source.
Everything Claude Code: 286 skills, 68 subagents, and it runs on Cursor, Codex and Gemini too. The setup won Anthropic's own hackathon.
https://t.co/SQ7MilNBCx
Claude Code writes too much code and you pay for every line.
Ponytail is a free plugin that makes it check the codebase first and reuse what already exists. It only builds when there's nothing to use. Works on any AI model.
https://t.co/QNs4BIsodw
If you can build the app, humanise the copy, wire up payments and publish to your own domain in an afternoon, what's actually stopping you shipping?
I don't think it's the tools anymore.
You don't need a nuclear warhead to open a door. I've asked Fable 5.1 how many calories are in my protein shake, so I'm as guilty as anyone.
Match the model to the job and your credits go a lot further for the same output.
Claude only reads a video's transcript. Gemini agentic video actually watches it.
Live in the API now: it skims the transcript, then pulls frames only where it matters. Up to 88% fewer tokens and 66% lower cost on a two-hour podcast.
https://t.co/h3yBlIF7od
Stop babysitting your agent every five minutes.
GPT 6 Astra Goal mode: type /goal, describe the finish line, and it runs its own planning loop until the outcome is done. A destination instead of directions.
https://t.co/J1kSWWNt12
One model auditing its own code is grading its own homework.
Claudex Loop is a free skill for Claude Code / Codex: Fable 5.1 writes the plan, GPT 6 Astra audits it, they swap roles and loop until both agree.
https://t.co/Cmg2o5U19i
Google just put its own Claude skills on GitHub for free.
google/skills is the official playbook for Google's own tools, tested by the people who built them. Drop them into Claude and it knows the playbook before you ask.
https://t.co/rWaiIXRZ91