Rapier

❗️UPDATE THREAD❗️ New updares with @MixedByCR & I’s music AI generator web-app. VERY big news… #AImusic #MusicTechnology #artificalintelligence #eacc

I think of this tweet every day. I can't stop myself. Every time I see anything YC related now, it immediately comes to mind. I hope to be free of this tweet someday.

@levelsio Sad I couldn’t 🫩

Okay and who of you couldn't reply to this? Please please only if you couldn't okay?

@bygregorr @Pirat_Nation AI slop response

@Pirat_Nation Building a product can be tough when public perception shifts quickly. I've seen my own app's reputation change overnight due to external factors. Reputation is fragile.

Sam Altman admitted OpenAI's rushed Pentagon (DoD/DoW) deal appeared "opportunistic and sloppy" amid strong backlash. > ChatGPT app uninstalls in the US: Surged 295% day-over-day > One-star reviews: Spiked sharply up 775% on Saturday in some reports

@nic_carter Boot licking monkey

If a top AI CEO in China told the CCP to go kick rocks when they asked for help, that CEO would be instantly sent to prison. This is the correct approach Letting AI CEOs play politics and dictate policy for the military and soon the entire country like their own personal fiefdoms is appalling and undemocratic If Trump doesnt bring Dario to heel now, we will simply end up completely subjugated by him and his lunatic EA buddies

@everestchris6 What’s the actual closing rate? Is pricing a static price you are working with? Or do you change it on certain industry

My OpenClaw bot runs 6 AI agents 24/7: - Finds local businesses without a website - Builds a custom demo site for them automatically - Sends outreach with the preview + payment link - Handles objections and closes the sale Most local businesses don't have a website, this system finds them, pitches them, and collects payment automatically Reply "OpenClaw" and I'll send you early access (must be following)

@vitddnv MRR

I just finished writing my most valuable PDF yet: "$0 to $10K MRR in 14 days" (19 pages) It's everything I wish I knew when starting out. I might charge for this in the future, but for now… Reply "MRR" and I’ll DM it to you for free (must follow)

@awsdevelopers AWS

Reply to this tweet with "AWS" and we’ll tell you which AWS Service you are

@weretuna This is so brilliant words don’t even describe how perfect it is

i hate talking to fake people so i made a button that forces you to say nigga when you try to book a call with me

@sama LMAO huge W

Peter Steinberger is joining OpenAI to drive the next generation of personal agents. He is a genius with a lot of amazing ideas about the future of very smart agents interacting with each other to do very useful things for people. We expect this will quickly become core to our product offerings. OpenClaw will live in a foundation as an open source project that OpenAI will continue to support. The future is going to be extremely multi-agent and it's important to us to support open source as part of that.

@kohya_tech Try Nvidia Audio Flamingo 3 with stage35 think adapter and llava runtime. It ends up providing much higher detail annotation. I’ll share comparison once I’m done. Running a Lora fine tune now

@AndrewTrapier It's a different audio model. ACE-Step captioner didn't work very well. huggingface.co/Qwen/Qwen2-Aud…

手元の50曲くらいで適当にACE-Step-1.5のLoRAを学習してみた。同一seed同一設定で、1つ目LoRAなし、2つ目LoRAあり。雰囲気出てるかな……。 2e-4で100 epoch、20分くらい。わりと早く過学習しちゃう。

@kohya_tech That’s the auto captioning the Ace Step provides right?

@AndrewTrapier All songs are captioned. Captioning was done by Qwen2-Audio.

@galileowilson @Legendaryy An arm and a leg

@Legendaryy what’s the price for opus 4.6?

$1 Per Hour AI Agents Just Dropped MiniMax M2.5 just shipped. Open source. 80.2% on SWE-Bench Verified. $1/hour for continuous agent operation at 100 tokens per second. A Chinese startup just made infinite-running AI agents economically viable for everyone. We've gotten used to a new model dropping every week with impressive numbers. So let me be specific about what makes M2.5 different from the noise. The cost structure changes everything. At $1 per hour with 100 tokens per second, you can run an AI agent continuously the way you'd run a cloud server. Not per-task. Not per-query. Continuously. An always-on coding agent that watches your CI pipeline, catches bugs, opens PRs, and fixes test failures. Running 24/7 for $720 a month. That's less than a junior dev's daily rate. MiniMax went after the three benchmarks that matter most for agent workflows: SWE-Bench (can it actually fix real bugs in real repos?), BrowseComp (can it search the web and find correct information?), and BFCL (can it call tools reliably?). 80.2%, 76.3%, and 76.8% respectively. 37% faster at complex tasks. Speed sounds like a nice-to-have until you're running an agent loop where the model calls itself 50+ times per task. The big picture: MiniMax isn't competing with Opus on creative writing or with GPT-5 on reasoning puzzles. They built a model specifically for the thing everyone actually needs in 2026: reliable, fast, cheap agents that run all day. Who is @MiniMax_AI? A few things stand out. MiniMax IPO'd just one month ago. Their stock doubled on day one. That's not a small company shipping side projects. That's a publicly traded company with real pressure to deliver. 70% of revenue comes from overseas. Unlike most Chinese AI labs that dominate domestically and struggle abroad, MiniMax's consumer product Talkie (an AI companion app, think Character.ai) has millions of users globally. They know how to ship products that non-Chinese users actually want. The MiHoYo connection is real. MiHoYo (the studio behind Genshin Impact) was MiniMax's angel investor. That's a gaming company with $5B+ annual revenue betting on AI infrastructure. Not a coincidence that MiniMax's models are optimized for real-time interaction loops. And Jensen Huang publicly called them out as "world-class." When the CEO of NVIDIA says that about a Chinese AI startup, the industry pays attention.

@kohya_tech When training did you associate that caption prompt to all the song data? Or did you only provide the audio file

In this example, I trained on songs with similar atmospheres. I think there is a difference especially at the end of the songs. When training on songs by the same singer, LoRA is more noticeable. However, overfitting seems to reduce the quality. I haven't yet found the optimal training settings.

@maxbittker @UnforcedBreath What discord?

@UnforcedBreath people on the discord are trying all sorts of stuff, including agent swarms

Running the server on x30 so the AI can test it's strategies faster (realtime screen recording)

@pleasekilljay 😂😭

nigga what am i reading

@NapoleonBonabot @grok is this real?

The greatest poster of all time

@sama Yesterday: "AI agents will do everything" Today: "You'll manage teams of AI agents" Turns out we're not being replaced. We're being promoted to robot HR.

Found it