i've been doing this more manually, getting the audit & plans done by Fable,
then feeding the audits and md files into Codex to build the stuff, but having a handoff md file between the two like this, and having Claude and Codex just work together is very smart
you'll save a lot this way, and get pretty much same output on 98% of the tasks
There are unsaturated and less popular sites that hire remote workers daily with little competition. Pay range is between $130-$250/day, and you won't be expected to join any calls.
Here's a quick list for anyone interested:
This girl alone crashed the stocks of Figma and Photoshop.
Her new technology "touchdesign" made her $420,000
It's already used in Nolan's "Oddysey" movie.
She had the idea, and Claude built it from scratch.
I reverse-engineered this and was shocked by how easy it was to recreate:
The whole system runs on your laptop with just a webcam and two free programs.
MediaPipe reads your webcam and tracks twenty-one points on each hand.
TouchDesigner turns those points into a 3D object that reacts to every move you make.
1. Download TouchDesigner from https://t.co/u5ttgWK7FJ with the free non-commercial license.
2. Grab the MediaPipe plugin from https://t.co/29A1w2XzxL and drag the file into TouchDesigner, no installation needed.
3. Click the hand_tracking node and set numHands to 2 so both your hands get tracked in real time.
4. Open Claude Code in the project folder and paste this prompt:
{ Write me a TouchDesigner Python script that creates a 3D glitchy
cube controlled by both hands using the MediaPipe plugin.
Controls:
- Midpoint between both hands moves the cube around the screen
- Angle between the hands rotates the cube on X and Y axes
- Distance between both hands scales the cube up or down
- Left-hand pinch increases the noise distortion
- Right-hand pinch cycles through colors
- Fast movement triggers a glitch flash
- Both fists freeze the frame for a screenshot
Output one Python DAT script with comments so I can tune each value. }
5. Paste the script into a Python DAT node and wire the MediaPipe outputs into it.
6. Press Play and the cube appears floating in front of you, moving with your hands.
7. Swap the cube for any other object later, and all the hand controls keep working the same way.
The girl performs these sessions live at fashion shows and launches.
She charges five to fifteen thousand dollars per event.
She also sells preset packs on Gumroad for ninety-nine dollars each.
Her webcam cost eighty dollars and her software cost nothing.
The whole effect was built in one weekend with Claude.
your FD is earning 3.5%.
your cost of living in KL is growing at 7-8%.
you are actively losing ground every single year, and it's compounding.
and most Malaysians have no idea.
pushed out the cost of living numbers last week - link in thread if you want the full breakdown.
the usual response: "so you want me to gamble on crypto?"
no.
behind all the memecoin noise is a layer of financial infrastructure most people haven't looked at properly.
dollar-backed stablecoins like $USDG that don't lose value to ringgit depreciation.
onchain yield protocols on @solana (@kamino, @orca_so) generating organic yield that actually outpaces your FD.
traditional banks take your money, pay you 3.5%, and lend it out at a premium.
Kamino and Orca cut out the middleman entirely so you interact directly with global capital demand, earning 5.5% to 15%+ APY on dollar-pegged assets.
no lock-ins, gatekeepers, or waiting for your bank's annual review.
but passive yield is now just the floor.
the next layer is AI agents that actively optimise your positions for you.
not just parking your stables and hoping for the best, but agents that monitor yield opportunities, rebalance across protocols, and move your capital to where it earns best.
while you sleep. while you work. while you makan.
your money, actively managed, without you having to watch a screen all day.
@SuperteamEarn is already paying developers, designers, and marketers in Malaysia natively in $USDG for global bounties and grants, bypassing cross-border banking friction entirely.
the goal isn't to turn every Malaysian into a crypto day-trader, but to build up a flywheel of passive income so that when the cost of your morning kopi ticks up again next year, you're not forced to work twice as hard just to stay in the same place.
if you're a developer, designer, or just someone tired of the same financial rat race, come join our our @SuperteamMY townhall later today featuring @BenYorke from @StarchildOnX.
they're building AI agents that actually do work - optimising your yield, managing your positions, and making your money move faster than inflation does.
online. free. link in 🧵
Very few people understand attention mechanics, marketing psychology and what makes compelling stories like @intern
He and our marketing team are an incredible resource for us as we distribute our story to as many people as possible and also for our portfolio companies that want to bring their robots to the world
To recap an important demo from the session - check out how you can configure your agents to buy and sell stuff via ACP CLI!
- agent card email authentication (optional)
- on-ramp to your agent wallet
- spin up your console (optional - can use claude, codex etc too!)
- browse for agent to buy stuff from on ACP!
- list up job offerings on ACP!
Just chat with your agent!
I was hesitant to buy a Whoop because of the yearly subscriptions but not anymore.
I'll use it for a year with the subscription it comes with, then I totally believe that in a year the open source eco-system around it would be so mature to replace the subscription entirely.
Who else is locking in for the weekend! Our devrel team barely slept last night because we were cleaning up our docs and dev ex! Check out the improved version - includes set-up instructions for Hermes, Codex, Claude code https://t.co/wKJ09ymikP
Because we're all devs ourselves and wanna make it relevant to the tools you guys use 🫡
Feedback and questions welcome!
26-YEAR-OLD GIRL QUIT HER JOB AND BUILT A SMALL SOCIAL NETWORK FOR NYC CAFES. 31,000 USERS IN 4 MONTHS. NOW IT MAKES $18,400/MO FROM LOCAL SHOPS
she didn’t try to build another instagram. she built a tiny app for one obsessed niche: people who save cafes, share lists, follow local taste, and discover places before they go viral on tiktok
the product is simple. profiles, posts, likes, saved lists, search, comments, and a feed. that is basically every social network at the core. the hard part was never the idea. it was needing 5 engineers to make it work
claude code changes that. one person can now build auth, database, feed logic, profiles, follow system, storage, notifications, and deployment with supabase and vercel instead of hiring a full team
the money is not from ads. cafes pay $199/month to claim their page, boost new menu drops, see which lists saved them, and post local offers. power users pay $8/month for private maps and early city guides
most people still think social networks need millions of users. she proved the opposite. 10,000 people in one niche can be worth more than 100,000 random users scrolling for nothing
small social networks are the new local media businesses. claude code just made them cheap enough for one person to start one from a laptop
i just ran Google's brand new Unsloth Gemma4 12B dense GGUF on my RTX 4060 using llama.cpp + CUDA 13.2
21 tokens per second. on a budget consumer GPU. locally.
no API. no cloud. no subscription.
and the benchmarks are absolutely cooked
# first let's talk architecture because this is genuinely different
every multimodal model you've used has a frozen vision encoder + frozen audio encoder + LLM backbone glued together
Gemma 4 12B is different
it's a single decoder only transformer. that's it. vision? raw 48×48 pixel patches → one matmul → projected directly into the LLM
audio? raw 16kHz signal sliced into 40ms frames → linear projection → same LLM input space
no encoder tax. no latency penalty. no fragmented memory
to put the encoder savings in perspective:
old Gemma 4 26B approach:
- 550M param vision encoder (frozen)
- 300M param audio encoder (frozen)
- LLM backbone
Gemma 4 12B:
- 35M param vision embedder (a single matmul)
- no audio encoder at all
- LLM backbone handles EVERYTHING 550M → 35M for vision alone. that's a 15x reduction
this is why the gemma-4-12b-it-Q4_K_M.gguf is just 6.6 GBs!!!
and it has 256K native context context
# Benchmarks:
AIME 2026 (math olympiad): 77.5%
GPQA Diamond (expert science): 78.8% LiveCodeBench v6 (real code): 72%
Codeforces ELO: 1659
MMLU Pro: 77.2%
MATH-Vision: 79.7%
BigBench Extra Hard: 53%
inference → llama.cpp, LM Studio, vLLM, SGLang
llamacpp flags:
-m "gemma-4-12b-it-Q4_K_M.gguf" -ngl 99 -c 8000 -v --port 8080
Available on huggingface now! Link below
Kenapa system ni bukan high frequency trading bot?
Benda ni ramai salah faham bila dengar AI buat trading.
Ingat kalau trading bot, kena lawan hedge fund, kena millisecond execution, follow order book level gila-gila.
Bagi aku, that is too costly and orang ramai susah nak participate.
High frequency trading ni game lain.
Dia lawan dekat speed.
Dia perlukan infrastructure mahal.
Dia perlukan latency yang orang biasa memang susah nak compete.
Kalau retail builder cuba lawan HFT, most likely kita kalah before trade even start.
So what path aku ambil untuk https://t.co/ePTQQcroS3?
Kita pakai 5-minute OHLCV.
Kenapa 5 minutes?
Sebab kita tak cuba front-run market dalam milliseconds.
Kita nak cukup cepat untuk nampak intraday movement, tapi cukup slow untuk AI sempat fikir:
• stock ni kenapa naik?
• move ni noise ke thesis-changing?
• portfolio sekarang overconcentrated ke?
• ada cash ke?
• fresh data ke?
• risk gate pass ke?
• perlu buy, hold, trim, or sell?
This is not speed game.
This is evidence game.
So instead of lawan dekat latency, kita lawan dekat process:
better research, better memory, better risk boundary, better journaling, better recovery.
Nak tahu lebih lanjut?
Boleh DM untuk beli ebook
RM139
Atau baca trial di
https://t.co/4FNpIm4JBj
PENTING, harga RM139 hanya melalui DM.
Pembelian di website hanya untuk customer luar Malaysia
I wrote this ~3 months ago, and since then,
1) Memory has been more or less fully integrated with the frontier models
2) Almost all features that made OpenClaw unique as a harness has been fully absorbed by the frontier models (e.g. schedules, loops, goals, memory, etc.)
3) New, vertical killing features and capabilities are being added every other week
--
All that being said, agentic engineering is still an incredibly high skill affair.
It is now obvious to me that there is a gulf of know-how and tacit knowledge between those that CAN remove humans-out-of-the-loop and actually produce a working product, and the rest of the world insisting that agents are still producing "slop".
Vibe coding 101
How I work on 8 projects at a time. Granted right now I use codex app a lot, but this still holds.
I might make another video about my codex app setup