Bhaje Vishesh Sundaram, Samast Paap Khandanam. 🙏🚩
B @grok
H
A
J
E
V
I
S
H
E
S
H
S
U
N
D
A
R
A
M
🙏🚩
S
A
M
A
S
T
P
A
A
P
K
H
A
N
D
A
N
A
M
🙏🚩
Swabhakta Chitta Ranjanam, Sadaiva Rama Advayam. 🙏🚩
S
W
A
B
H
A
K
T
A
C
H
I
T
T
A
R
A
N
J
A
N
A
M
🙏🚩
S
A
D
A
I
V
A
R
A
M
A
A
D
V
A
Y
A
M
🙏🚩
भजे विशेषसुन्दरं समस्तपापकण्डनम्।
स्वभक्तचित्तरञ्जनं सदैव राममद्वयम्॥
The cat knocked something off the table and walked away and they looked so proud of themselves Seeing a field of wildflowers is the closest thing to magic I've ever seen
This graph made him $19,700.
Not because it looked cool.
Because it remembered everything.
The creator connected Claude Code to Obsidian.
Every note became searchable.
Every project stayed connected.
Every conversation became reusable.
Cost:
- $20/month.
Saved:
- 20 minutes every client email.
Over a year...
That's hundreds of hours.
AI gets smarter when your knowledge compounds.
Most people keep starting over.
Save this.
Claude is currently experiencing a major outage...
boris cherny, the engineer behind claude code, quietly built a context system that survives even when claude doesn't and almost nobody noticed
here's the problem he solved...
most people don't have a claude workflow. they have one long conversation holding the goal, files, decisions and unfinished work together until the model disappears and the entire system disappears with it
the context manifest keeps the goal, constraints, active files and source priority outside the chat, so claude is never the only place that understands the task
every important decision enters a continuity log the moment it happens, including what changed, what was rejected and what still needs proof
if claude goes down, the same context packet opens in another model... when claude returns, it reads the updated state and continues without rebuilding anything
the model can disappear... the context layer is the part that keeps compounding
full context engineering breakdown is in the article below