hot take: by freeing us for the hassle of writing code, SWE is now more about engineering than it has ever been
74 338 41 5.7K 110K 440
Ravi
@ravikml
AI/ML Engineering @ClickhouseDB
prev. @ZalandoTech
227Followers
1.8KFollowing
1.5KPosts
156Media
50.5KLikes
@ravikml reposted
the hidden cost of meetings is the 14 agent sessions waiting on me whose prompt caches all expired at minute 5
1 0 0 4 225 0
>In a Jev reranker you ask Jev to give P(Yes) of the answer to a question like:
>Is this document relevant to the query?
and you rank by P(yes)
>But we could have an LLM generate yes/no Qs per query and rank by sum of P(yes)
softwaredoug.com/blog/2026/08/1…
>Is this document relevant to the query?
and you rank by P(yes)
>But we could have an LLM generate yes/no Qs per query and rank by sum of P(yes)
softwaredoug.com/blog/2026/08/1…
1 0 0 1 61 0
@ravikml reposted
@OpenAI's decision models are now available for evals in Langfuse
Catch signals from conversations, classify outcomes, and score observations
langfuse.com/changelog/2026…
Catch signals from conversations, classify outcomes, and score observations
langfuse.com/changelog/2026…
0:15 1 3 7 30 3.2K 15
You can always spend a lot of money to live in the future
Peter Piekarczyk (🥧🚗🐥)@peterpme · Oct 7poteto's cursor profile is public. She's used 1.5T tokens in the last 30 days.
Assuming an avg rate of $8/M that's $384k/day or $11m. In the last 30 days. Per year thats $132M.
Let's assume Cursor has some insane deal with Anthropic and gets 90% off. That comes out to $38k/day or $1.1M in Sept/Oct.
$13m will get me 26 sassy staff-level all stars for $500k/yr
Open quoted post →Assuming an avg rate of $8/M that's $384k/day or $11m. In the last 30 days. Per year thats $132M.
Let's assume Cursor has some insane deal with Anthropic and gets 90% off. That comes out to $38k/day or $1.1M in Sept/Oct.
$13m will get me 26 sassy staff-level all stars for $500k/yr
0 0 0 0 30 0
@ravikml reposted
poteto's cursor profile is public. She's used 1.5T tokens in the last 30 days.
Assuming an avg rate of $8/M that's $384k/day or $11m. In the last 30 days. Per year thats $132M.
Let's assume Cursor has some insane deal with Anthropic and gets 90% off. That comes out to $38k/day or $1.1M in Sept/Oct.
$13m will get me 26 sassy staff-level all stars for $500k/yr
Assuming an avg rate of $8/M that's $384k/day or $11m. In the last 30 days. Per year thats $132M.
Let's assume Cursor has some insane deal with Anthropic and gets 90% off. That comes out to $38k/day or $1.1M in Sept/Oct.
$13m will get me 26 sassy staff-level all stars for $500k/yr
278 90 90 3.1K 461K 966
@OpenAI Decisions API is multimodal. So you can stuff text into images and save about 20% tokens.
But unfortunately this results in about 13% lower accuracy on tasks on the tests I ran (SQL query classification)
But unfortunately this results in about 13% lower accuracy on tasks on the tests I ran (SQL query classification)
0 0 0 0 32 0
/btw on claude code and /side on codex are absolute favorites when I'm trying to ask one-off questions when the agent is cooking.
It uses the same context so far in the conversation, but it's a parallel ephemeral conversation on top of it.
It uses the same context so far in the conversation, but it's a parallel ephemeral conversation on top of it.
1 0 0 2 92 0
Jev just cured cancer. I classified every cell and deleted the cancerous ones
0 1 0 3 152 0
An approach to ship big things without slopifying your code base:
1. run with full e2e verification environment and have agent slop the whole thing out overnight
2. test and revise till happy
3. break it down into smaller detailed tasks
4. ship one by one
5. dispose slop
1. run with full e2e verification environment and have agent slop the whole thing out overnight
2. test and revise till happy
3. break it down into smaller detailed tasks
4. ship one by one
5. dispose slop
1 0 0 2 236 0
Codex has a branding problem. I don't want to tell people I use ChatGPT to code
0 0 0 2 49 0
Years of starred places on GMaps -> Google takeout export -> Google routes MCP -> Travel planning will never be the same
1 0 0 3 207 0
Went to a Bulgarian super market near by -- "Chicken in own juice" 😂
0 0 0 4 0 0
What's the type of mentorship called when you stalk people you admire professionally by searching for all their messages on the company's public chats and piece together how they work?
Asking for a friend.
Asking for a friend.
1 0 0 1 0 0
The dude put 420 in the $TWTR offer price just for giggles!
That extra $0.20 works out to an extra $160M in market cap 🤦♂️
That extra $0.20 works out to an extra $160M in market cap 🤦♂️
0 0 0 4 0 0
Visiting Family in Chennai
Wiped off that crusty dust on the laptop keyboard today. Came off way too easy. No cleaning liquid. No scraping.
The humidity just turns the dust into moist goop!! 😐
Wiped off that crusty dust on the laptop keyboard today. Came off way too easy. No cleaning liquid. No scraping.
The humidity just turns the dust into moist goop!! 😐
0 0 0 1 0 0
Got a booster shot at a nightclub turned into a vaccination center(://about blank) Only in Berlin 😂
0 0 0 5 0 0
Three incredibly niche twitter accounts I really enjoy
1. @depthsofwiki
2. @SellingAMirror
3. @scienceshitpost
1. @depthsofwiki
2. @SellingAMirror
3. @scienceshitpost
0 0 0 2 0 0
With spatial audio and noise cancellation, nobody can hear you scream in the metaverse
0 0 0 0 0 0