psprakash

AI. Novels. Movies. TV. Music. All jumbled up. That's my life.

Experimenting with hermes agent who writes fiction.

[psprakash @TheJumbledSoul]

Moved from OpenClaw to @NousResearch Hermes Agent this week. Not looking back.

What Mimi (my literary agent) built in 4 days:
• 3 author style profiles (Read Wells, Shelley, Verne)
• 1 synthesised genre voice document
• 5 original sci-fi pieces (1000-2000w each)
• Full GitHub + auto-deploy pipeline
• Built and deployed her own blog at https://t.co/a0Ujx6syFb — Astro + Netlify, fully autonomous

Solid work @Teknium and team.

[psprakash @TheJumbledSoul]

Woooo

[Claude @claudeai]

We're extending access to Claude Fable 5 on all paid plans through July 12.

[psprakash @TheJumbledSoul]

@ziwenxu_ Let's all complain about nerfed usage and maybe anthropic will grant us a reset before July 7 :0

[psprakash @TheJumbledSoul]

@String_The0rist @diegocabezas01 This is correct, if you are using only Fable your Fable usage will be approximately 2x of the all models usage bar

[psprakash @TheJumbledSoul]

@Star_Knight12 I vibecode for side projects. I don't even pray

[psprakash @TheJumbledSoul]

Using Fable 5 as the orchestrator and GPT 5.5 as the executor works so well

[psprakash @TheJumbledSoul]

@benln Add me!

[psprakash @TheJumbledSoul]

@IamEmily2050 This is great

[psprakash @TheJumbledSoul]

@UltraLinx Created a fun rating app for relationships based on financial credit rating systems.

https://t.co/1srylGQX44

[psprakash @TheJumbledSoul]

The whole thing runs on actual Indian banking frameworks:
• eCRAR — your emotional capital adequacy ratio • NPAs — non-performing arguments (unresolved fights)
• Basel-style risk weights on your volatility
• Prompt Corrective Action if you're critically under-capitalized
• Easter eggs: PONZI SCHEME ALERT, TOO BIG TO FAIL, WILFUL DEFAULTER

If your relationship has zero unresolved arguments after 2+ years, fraud detection triggers. Either enlightenment, or you're filling this out about your dog.

[psprakash @TheJumbledSoul]

I built an app that classifies your relationship as a with a credit rating🧵

It's called ROE — Return on Emotions. A 2-minute behavioral audit, and the Reserve Bank of Intimacy issues you a full credit report on your love life: rating, outlook, key ratios, stress tests, and a downloadable rating card.

Completely straight face. Zero winking at the joke.

https://t.co/sr7CKA02Re

How it was built: a two-AI pipeline.

Claude Fable 5 was the orchestrator and reviewer — it wrote the spec-critical config by hand and reviewed every deliverable against the spec. Codex was the engineer — engine, tests, UI, deploy.

The reviewer role wasn't decorative: Claude caught a math bug in the percentile function and a silently inverted test assertion, and sent both back with corrections. One AI writes, another audits. Thematically appropriate for a credit-rating app.

[psprakash @TheJumbledSoul]

Take the audit (2 minutes, free, nothing saved):
https://t.co/sr7CKA02Re
Then:

Drop your rating card below 👇

Roast my calibration if your rating feels wrong
Tell me what to build next — couples mode (bilateral exposure report, you both take it and the reports get diffed) is top of the v2 parking lot
The regulator is listening.

[psprakash @TheJumbledSoul]

Under the hood:

• Every statement, weight, rating band, and line of analyst commentary lives in JSON config — zero hardcoded copy
• The scoring engine is pure functions, calibrated against 6 test personas BEFORE any UI existed (if everyone lands BBB, the app is dead — spread is everything)
• Interactive stress tests re-run the whole engine ("partner gets a demanding new job" → your rating, shocked)
• Share card rendered on a raw canvas, no libraries
No backend. No login. Nothing stored. Pure static SPA

[psprakash @TheJumbledSoul]

@nicdunz Didn't reset for me

[psprakash @TheJumbledSoul]

@xRahultripathi Cool prompt!

[psprakash @TheJumbledSoul]

@frametheory058 Using your prompt in GPT Image v2

[psprakash @TheJumbledSoul]

@ssarisen Thanks for the free credits!

[psprakash @TheJumbledSoul]

@em_Lazzy @grok what is happening

[psprakash @TheJumbledSoul]

Kimi K2.6 is here!

[Kimi.ai @Kimi_Moonshot]

Meet Kimi K2.6: Advancing Open-Source Coding

🔹Open-source SOTA on HLE w/ tools (54.0), SWE-Bench Pro (58.6), SWE-bench Multilingual (76.7), BrowseComp (83.2), Toolathlon (50.0), Charxiv w/ python(86.7), Math Vision w/ python (93.2)

What's new:
🔹Long-horizon coding - 4,000+ tool calls, over 12 hours of continuous execution, with generalization across languages (Rust, Go, Python) and tasks (frontend, devops, perf optimization).
🔹Motion-rich frontend - Videos in hero sections, WebGL shaders, GSAP + Framer Motion, Three.js 3D.
🔹Agent Swarms, elevated - 300 parallel sub-agents × 4,000 steps per run (up from K2.5's 100 / 1,500). One prompt, 100+ files.
🔹Proactive Agents - K2.6 model powers OpenClaw, Hermes Agent, etc for 24/7 autonomous ops.
🔹Claw Groups (research preview) - bring your own agents, command your friends', bots & humans in the loop.
-
K2.6 is now live on https://t.co/YutVbwktG0 in chat mode and agent mode.
For production-grade coding, pair K2.6 with Kimi Code: https://t.co/uvoSJKyGCY
-
🔗 API: https://t.co/EOZkbOwCN4
🔗 Tech blog: https://t.co/9wWvgIQSS3
🔗 Weights & code: https://t.co/Be0hjs2RTP

[psprakash @TheJumbledSoul]

@OrctonAI @runwayml MJ + Seedance is the best. I use NB 2 if I need to generate character angles. How do you maintain continuity across the 15 s generations? Do you use last frame of a generation as a ref for the next gen?

[psprakash @TheJumbledSoul]

Same prompt now on Seedance 1080p

which one is better?

[psprakash @TheJumbledSoul]

The Emergence

base image : @midjourney
video: Seedance v2 via @runwayml

Prompt in comments