paul lodder

i build agentic AI systems and TB-scale weather data infrastructure

co-founder

[paul lodder @paul__lodder]

crazy that this was not even a year ago

[vas @vasuman]

Claude 4 just refactored my entire codebase in one call.

25 tool invocations. 3,000+ new lines. 12 brand new files.

It modularized everything. Broke up monoliths. Cleaned up spaghetti.

None of it worked.
But boy was it beautiful.

[paul lodder @paul__lodder]

@AndrewDeJackson @uptimerobot thanks, i just migrated to betterstack, but will keep pulsetic in mind if betterstack disappoints as well

[paul lodder @paul__lodder]

is it me or is @uptimerobot surprisingly flaky for a status monitoring service?

[paul lodder @paul__lodder]

@SimonNom1 @claudeai termux + vps you ssh into works quite alright

[paul lodder @paul__lodder]

a pattern i use with claude code/codex a lot:

have it write a one-off script to diagnose, reproduce, and fix an error/bug

and only then apply it into the actual codebase

it can iterate in isolation which reduces the cognitive load for the agent

[paul lodder @paul__lodder]

(me coping with my ever-growing dependency on llms to write code for me)

[paul lodder @paul__lodder]

i wonder: when cars were invented, were people worried they'd forget how to ride a horse?

[paul lodder @paul__lodder]

i wonder how many codex features are not implemented out of fear of copycatting claude code too much

like not being able to use C-n/C-p to move up/down a list...

why do i have to use my arrow keys???

[paul lodder @paul__lodder]

what a time to build

[paul lodder @paul__lodder]

depending on hyperscale cloud provider (aws, azure gcp) to host services is unnecessarily complex for the majority of use cases

just rent a server and host it yourself

[void. @iBuild]

a tech opinion you can defend like this?

[paul lodder @paul__lodder]

@rickasaurus good point and i think it's context dependent

with adapters for instance, removing the right degrees of freedom and defining the protocol that each adapter needs to adhere to feels crucial to scale that to any reasonable number

[paul lodder @paul__lodder]

this captures so well what it's like to stumble upon *the right* abstraction

[François Chollet @fchollet]

You haven't so much invented such a solution, as you have discovered the coordinate system in which the problem becomes trivial.

[paul lodder @paul__lodder]

a $5 Hetzner box + systemctl restart is 90% of what most startups need

[paul lodder @paul__lodder]

cloud is a psy-op making you believe bare metal is rocket science

[paul lodder @paul__lodder]

@DSPyOSS but why would the data be too underspecified for this but not for prompt optimization? cause the problem representation space is so much bigger?

[paul lodder @paul__lodder]

working with @DSPyOSS shifts the focus from prompt engineering to problem representation

LLMs are semantic machines. even changing a field name can materially affect results because you're changing the semantic relationship the model sees

[paul lodder @paul__lodder]

@DSPyOSS yeah so it's a bit of a stretch

it feels akin to differentiable neural network architectures

in the is_spam example, the ultimate output is is_spam, but there are many paths to get there

[paul lodder @paul__lodder]

@DSPyOSS exactly. i guess you'd almost want a meta-optimizer that searches across different problem representations and post-processing configurations to find what optimizes for your end goal

[paul lodder @paul__lodder]

@DSPyOSS for instance, this pattern has made a big impact for me:

instead of is_spam: bool

classify as promotional/phishing/personal/automated
then return category in [promotional, phishing]

gives the model richer semantic structure even though you only need the binary result

[paul lodder @paul__lodder]

@DSPyOSS this tripped me up initially. dspy resembles classical ml approaches, so i assumed labeled data was enough.

but problem representation still matters enormously