
Growing personal brands for ambitious Founders & CEOs on X ✨HIGGS-I31W8 Open for Paid Promotions. 📩 DM or 📧 [email protected]
California, USA
Joined May 2022
- Tweets164K
- Following516
- Followers15.9K
- Likes124K
Pinned Tweet
Open-source AI keeps getting stronger, and I’m all for it. 🚀
MiniCPM-OCR 1.0 (1.3B) is pushing document parsing and text recognition forward with impressive benchmark results.
→ 93.68 on Real5-OmniDocBench
→ 54.7 on OCRBench v2
→ 1.49× the throughput of PaddleOCR-VL-1.6
This is exactly why I believe open-source AI is the future.
My prediction remains the same: in the medium to long term, AGI could become accessible to everyone, running locally at little to no cost.
Powerful AI shouldn't be limited to a few. 🔓
Your OCR says it’s done—but did it get everything right? 🤔 In high-stakes workflows, even a small error can be costly.
🚀 Meet the MiniCPM-OCR suite: an open-source solution to parse fast and guard the results.
Two new models:
🔹 MiniCPM-OCR 1.0 — SOTA OCR performance and speed at 1B scale. Compact, accurate, and fast.
🔹 MiniCPM-OCR Jev — A 2B decision model that helps assess parsing accuracy, with performance approaching that of Qwen3.5-397B-A17B on JevOCRBench.
⚡ What’s new?
Small model, SOTA results: MiniCPM-OCR 1.0 scores 93.68 on Real5-OmniDocBench and 54.7 on OCRBench v2.
A sharp goalkeeper 🛡️: MiniCPM-OCR Jev is model-agnostic, checking text, formulas, and tables from different OCR parsers. It achieves 93.6 on JevOCRBench, outperforming JEV-35B-A3B (90.1) and approaching Qwen3.5-397B-A17B (93.8).
Up to 128 pages per minute 🤯: With DSpark, MiniCPM-OCR runs at 1.49× the speed of PaddleOCR-VL-1.6 and 3.62× that of MinerU2.5 Pro. The SDK’s concurrency optimizations further boost throughput to as much as 128 pages per minute.
Deploy where you work: Day-0 support for Arm, Houmo AI, Intel, and NVIDIA, plus 10 more chip vendors through FlagOS.
Ready for your stack: vLLM, SGLang, llama.cpp, and Ollama, with quantized models, fine-tuning support, and Agent Skills.
🔗 GitHub: github.com/OpenBMB/MiniCPM-O…
🤗 Hugging Face:
MiniCPM-OCR 1.0:huggingface.co/openbmb/MiniC…
MiniCPM-OCR Jev:huggingface.co/openbmb/MiniC…
👉 Try the Demo:platform.modelbest.ai/consol…
Use the API:platform.modelbest.ai/consol…
SafwonWith_AI retweeted
Have you picked up Phantom Blade Zero?
Check out how it looks on Wizstar!
Claim your free credits in the threads 👇
#Wizstar #PhantomBladeZero #aivideo
SafwonWith_AI retweeted
GEMINI JUST CHANGED THE GAME.
Google is pushing Gemini beyond simple chat with stronger agentic workflows, coding and multi-step tasks.
Most people still use it for answers.
Here are 8 prompts to make it work like an Al operator.
SafwonWith_AI retweeted
The Chinese community has reverse-engineered Claude's account banning mechanism. Here's what they've discovered:
Claude uses a points system. Accounts receive an initial score upon registration, then gain or lose points based on behavior. Falling below the threshold results in a direct ban. Recovering points is very difficult.
Factors affecting scores:
Registration
> A clean IP is the most important factor. It doesn't have to be residential broadband, but it must be clean.
> Gmail is no longer the optimal choice; there are email services with higher weight.
> Accounts that prompt for KYC right upon registration can be discarded.
> Registering via the mobile app adds points.
Payment
> Subscribing immediately after registration deducts points.
> Jumping straight to Max deducts a large amount of points. It's best to wait two days and subscribe to Pro first.
> Being prompted for KYC during payment indicates the account is in a review period, but it won't necessarily get banned.
> Foreign entity bank cards add points. Google Play works too. Virtual cards deduct points.
Usage
> Don't exhaust the limit in one go. It's better to rotate between two accounts than to max out one.
> The desktop client gives a small points boost. Use the web version as little as possible.
> The mobile app also gives a small points boost.
> Enabling the memory feature adds points. The more memory files, the higher the score.
IP and Device
> Stability is key. Don't switch IPs; stick to the same one as much as possible.
> Proxies must be in global mode with TUN enabled.
> Logging in from multiple devices deducts a large amount of points. The ideal setup is one computer, one client.
> Frequent IP changes, frequent logins, and high concurrency all deduct a large amount of points.
Account Age
> Old accounts have huge leeway for errors. 2023 accounts clearly score higher.
> Scores grow over time, but very slowly.
Other
> No difference between Windows and macOS. Linux deducts points.
> Time zone and language have basically no impact. You won't get banned just for using Beijing time.
> Telling Claude you're Chinese is fine; memory won't trigger a ban.
> Heavy use of Fable deducts points.
> Upgrading to Max likely puts you in an observation period, unless it's a high-weight old account.
Observation Period and Banning
> When nearing the threshold, the account may enter an observation period. A typical sign is getting logged out.
> Continuing to deduct points during the observation period leads to a ban.
> High-score accounts can survive multiple observation periods.
Follow me to learn how the services you use every day really work behind the scenes.
Follow @rakib_hossen_ai for more content like this
Stop treating every new creative idea like a new project. Explore Flova 1.5. flova.ai/?refCode=JMCJVVYT
🤝 Paid partnership
Flova 1.5 | One Chat with Multiple Skills
One project. Multiple Skills.
Bring different Skills together in one creation!
- You set the goal. Agent plans the workflow.
- Mix and match Skills to explore different stories and styles.
- Refine your creations and take your ideas even further.
✨ Mix and Match Skills. Turn your ideas into videos, your way!
Flova 1.5 | One Chat with Multiple Skills
One project. Multiple Skills.
Bring different Skills together in one creation!
- You set the goal. Agent plans the workflow.
- Mix and match Skills to explore different stories and styles.
- Refine your creations and take your ideas even further.
✨ Mix and Match Skills. Turn your ideas into videos, your way!
Introducing Cartesia Managed Agents: our end-to-end voice agent platform for enterprise.
#1 on @ArtificialAnlys Speech Agent Arena and @ServiceNow EVA-Bench, built by co-designing our voice models and harness as one system.
Check it out → cartesia.ai/blog/managed-age…
Most voice datasets lean on a small group of heavy contributors. Good to see a release that clearly doesn't.
🤝 Paid partnership
SafwonWith_AI retweeted
Drop 1/15: Introducing Joy, a dual-channel conversational dataset.
Real, unscripted conversations between two people, each on their own track. Built to teach voice models how people actually talk: when to listen, when to jump in, when to say “hmm”.
100,000 hours.
50,000 speakers.
10 languages.
SafwonWith_AI retweeted
Oxide just raised $445M at a $6B valuation.
@bcantrill and @sdtuck started @oxidecomputer to build the computers the hyperscalers keep for themselves. @EclipseVentures led the seed when nobody else would. now they're profitable, and he says they didn't even need to raise.
i also asked him if AI is a bubble, why america hates data centers, and why every oxide employee makes the same salary.
nontechnical, episode 4.
00:00 $445M at $6B, and they didn't need it
03:06 one wire transfer bigger than their whole seed
07:14 "they don't have the guts. they're too chicken."
09:09 100% banked at SVB
13:04 a 92-year-old kicking their asses in board meetings
14:48 "would you like me to name names?"
16:58 a16z loved it, then sent the first rejection
20:19 leaving a first round partner meeting for the A's
22:52 "Marc Andreessen was offended"
23:05 NEA: "we will only invest in software"
25:55 "legends are dead"
28:32 "when a VC says 'I've got a technical question'"
42:39 the firm that walked away from a signed term sheet
46:48 "if I could get sequoia, do you think you'd be here?"
49:41 "they were cowards"
52:36 the nuke email he wrote and never sent
58:33 four years later, they asked for a reference
1:02:01 "politics and business don't mix"
1:10:51 is AI a bubble?
1:14:56 how he ruined a dinner party in 1997
1:20:44 "people hate data centers"
1:22:08 how sun got thrown through the windshield in 2000
1:26:26 why everyone at oxide makes the same salary
1:30:58 the worst hire of all time
1:40:42 when oxide goes public
SafwonWith_AI retweeted
"Sorry for the late reply" makes you look junior.
Senior leaders never say it.
Here are 6 things they say instead:
SafwonWith_AI retweeted
My boss borrowed my iPhone for "2 minutes".
He's ex-FBI cyber.
He gave it back and whispered: "Bro... turn off #3 right now. They're tracking everything."
Apple won't tell you this.
5 settings in iOS 26 that are spying on you:1. Stop Apple recording you
That "Hey Siri" you never use?
Apple saves it and workers listen to "improve Siri".
Delete it.
Settings → Privacy → Analytics → Turn OFF "Improve Siri & Dictation"
They don't need your voice.
SafwonWith_AI retweeted
My boss borrowed my iPhone for "2 minutes".
He's ex-FBI cyber.
He gave it back and whispered: "Bro... turn off #3 right now. They're tracking everything."
Apple won't tell you this.
5 settings in iOS 26 that are spying on you:
Producers, The Search wants a jingle and an ad and there's a share of $100K on the line. 5 days left. This one's for us
🤝 Paid partnership
SafwonWith_AI retweeted
I'm a one-person account. My org chart got a second box anyway.
I test AI tools and turn them into tutorials. Every tutorial has numbers in it, and for a long time the person checking them was me, at midnight, tired.
So the org chart looks like this now.
Me: testing, recording, writing.
Viktor, my AI employee in Slack: everything that has to be true before a post goes out.
Fact-check: the first night, he looked for a source on all 26 numbers in 3 threads. 4 flagged with no source, 2 quoted wrong.
Morning desk: his notes in Slack by 8:30, every day. Around 10 errors a week caught before I publish.
Archive: when an app I'd covered shipped an update, he flagged my 3 old posts about it and drafted a correction for each. Nobody asked.
That's about 7 hours a week back, and every call is still mine. He changes nothing in a thread. I cut or keep.
A one-person account still needs a fact-checker. Mine is on the org chart now.
Hire @viktor_com for your team. Start free.
#AIemployee #AItools
Paid Partnership
🤝 Paid partnership
SafwonWith_AI retweeted
🤚 STOP telling Claude, “Act like a marketing expert.”
Here’s how to make Claude break down HORMOZI’S $100M strategy.
🔥 Use these 7 prompts to uncover the marketing principles behind his biggest ideas:
SafwonWith_AI retweeted
For repetitive coding work, I'd rather describe the task clearly and review the result than spend time repeating the same manual steps. That's the kind of workflow I'd explore with Qoder.
DeepSeek-Flash is free through Oct 31.
Try Qoder here: qoder.com/qoder
🤝 Paid partnership
SafwonWith_AI retweeted
Didn’t expect this part to be the thing that caught my attention:
Sunra Enterprise deals can go up to 40% off leading models’ own API prices, depending on volume.
If you’re generating at serious scale, that’s not a tiny optimization.
That’s something worth running the numbers on.
SafwonWith_AI retweeted
One use case I'd explore with Qoder is building a small feature from a clear task description, then checking the finished files to see whether the agent got it right.
DeepSeek-Flash is free in Qoder through Oct 31.
Download it here: qoder.com/qoder
🤝 Paid partnership
SafwonWith_AI retweeted
Who said product ads have to be predictable? 📞🪟 Take a simple product and give it a story worth watching.
🔥 Dreamina’s first month costs just $0.29. From Oct 20–30, save 60% on CapCut Pro or 40% on CapCut Ultra on CapCut Desktop only.
Create your visuals with Dreamina, then edit in CapCut PC. Juggling multiple clips, sound effects, and music? Use CapCut PC Multi-Track Editing to bring every element together into one seamless ad. 🎬
Each product requires a separate subscription to enjoy its respective offer.
CapCut Sale: tinyurl.com/2v23rn22
Dreamina + CapCut Offer: tinyurl.com/5xbtv4x6
#CapCutDidThat #Dreamina #Smallbusinessmarketing #CapCutSale
🤝 Paid partnership




























