shiddiq retweeted
ok this is the part where agents get f*cking real
Karpathy estimates only 0.2% of people have watched an agent finish weeks of work from a single prompt
one /goal in Claude Code puts you in that 0.2%
the other 99.8% haven't seen it yet
the creator of Claude Code gives the model three things for long jobs:
-> the goal
-> how to measure it
-> how hard to try
write those into /goal and Opus 5.5 changes one thing, measures it, then keeps or reverts
after every run it redraws a progress chart, so one look in the morning tells you how far is left
goals worth trying:
> the homepage loads in under 2 seconds on a phone
> no page scrolls sideways at phone width
> the build finishes in under a minute
one rule before you leave it alone: never delete features or tests to hit the number
the person who hands Opus a measurable goal at night walks into the morning standup with finished work
pair it with an Obsidian second brain and the agent remembers why every change was made
that build is in the article below ↓
shiddiq retweeted
Been working on setting up my autonomous research team for Flamingo Compute. Made a system diagram to show how I'm setting things up.
Here's what it has:
> a coordinator that routes work to agents, reviews code changes, and runs the hardware tests
> a literature scout that sources and reads research and proposes experiments based on prior art
> a protocol designer that writes the experiment plan and hypotheses
> a research critic that checks every citation and statistic, and sends the work back if the evidence is weak
> a results analyst that runs the stats from our experiments
> a grants scout that hunts for the grants, fellowships and partners that fit what we're building
> A Content Bot that translates research results into post ideas
> coding agents that write the code and give every PR an AI reviewer
> a database and an experiment tracker that keep the receipts for every protocol, run, and result
> a raspberry Pi 5 running a small local model off an Anker SOLIX solar battery to validate variable energy compute models
Already got some experiments running. Looking forward to showing data soon!
shiddiq retweeted
Most Gen Z & Millennial Indonesian Elites will never be as successful as their parents.
Jakarta Elites live in a tight bubble that secludes them from mass Indo consumers.
Their speak English not Bahasa, attend international schools, and live in Jaksel.
Hence, any business idea they have tends to only serve people like them:
Padel courts, Senopati restaurants, or mid-tier fashion brands.
However, Indonesia is a bottom-of-the-pyramid economy where only businesses that serve the mass market can get huge ($100m revenue).
But they can’t serve the mass market because they rarely interact with them.
They're out of touch from the rest of Indonesia -- unlike their parents/grandparents, who grew up poor and got rich by serving the masses.
In Indo, having a privileged life means you lack the experiences that made your family successful in the first place.
Apa kabar kawan ? sesudah tarawih ini mau ngopi sejenak ?
kalo belum sempet cek toko cek link dibawah yaa
malabarroastery.com/link-tre… – at Bojongsoang, Indonesia
26
Bukan cuma soal rasa, tapi juga kesegaran. Kami menyediakan stok fresh mulai dari Arabica Sunda Chil yang aromatik sampai Robusta kiloan yang mantap untuk campuran es kopi susu kekinian.
📍 Mampir ke Roastery Kami:Jl. Jupiter Utama No. E2/106, Sekejati, Kec. Buahbatu, Bandung – at Rancasari, Indonesia
250
shiddiq retweeted
“It is true that a mathematician who is not somewhat of a poet, will never be a perfect mathematician.”
- Karl Weierstrass (1815 - 1897)
Nilai plus besar: sudah punya customer atau PO HoReCa existing. Full lapangan, mobile, target oriented. Minat? linkedin.com/jobs/view/43453…
26
















