Cindy Sridharan @copyconstruct

copyconstruct.medium.com San Francisco, CA Joined July 2010

Tweets

12K
Followers

54K
Following

160
Likes

36K

Cindy Sridharan @copyconstruct

2 hours ago

@pantasw2 @manasque It has both open and closed source test suites. SQLite is a library with multiple layers; all but one of the layers has oss test suites, iirc. There’s plenty of great ideas in the open source test suites, covering a host of domains.

0 0 0 84 0

View Details

Cindy Sridharan @copyconstruct

21 hours ago

To create my claude.md, I had it read: - my favorite programming books + blogs - the best oss projects - the code of the engineers I respect the most Claude (when given proper instructions, constraints, and context) now writes better code than me, almost always.

zack @zack_overflow

a day ago

the coding agent haters are getting dunning-kruger'd so hard right now do you really think that your engineering ability is better than the models at this point? it is very very likely that fable is better than you in every branch of software it has been trained on all the

254 127 2K 281K 479

84 163 4K 280K 4K

View Details

Cindy Sridharan @copyconstruct

4 hours ago

@dmitriid If that was your takeaway from that article, there’s no point engaging with you in good faith anymore. 👋

1 0 0 78 0

View Details

Cindy Sridharan @copyconstruct

5 hours ago

I’m literally that weirdo who wrote (all by hand) a 15K word blog post on: - why one cant treat models as oracles - how models gaslight you all the time - what it takes to build confidence in the model’s output copyconstruct.medium.com/how-to-effecti… So yeah sure, I believe in magical woo-woo

Dmitrii @dmitriid

18 hours ago

@copyconstruct I don't expect models to do things well automatically. I'm just pointing out that you believe in the magic too much. Including things "I gave it things to read" lol

3 0 10 10K 1

7 18 310 14K 290

View Details

Cindy Sridharan @copyconstruct

18 hours ago

@dmitriid None of this is true IME. People who expect models to automatically do the right thing are bound to be disappointed. Give it good instructions, constraints, context and multiply verification by tenfold, and you get way better results than any human can achieve on their own.

3 0 46 2K 6

View Details

Cindy Sridharan @copyconstruct

18 hours ago

@dmitriid Yes, and that’s why I have pre-commit review agents. A swarm of them runs prior to every commit to each verify a separate design principle or guidance.

Cindy Sridharan @copyconstruct

2 days ago

Many moons ago, I wrote about “post-commit” reviews. In the agentic era, a workflow I find extremely useful is … *pre-commit* agentic reviews. For large changes, the review agents take 10x the time it takes the coding agent to create a patch, but is totally worth it, IMO.

18 5 131 23K 114

2 2 88 13K 62

View Details

Cindy Sridharan @copyconstruct

19 hours ago

@sidharthkmenon The edge cases are all handled comprehensively, it always tries at least two prototypes if not more before settling on the best design, which I never did, it has very rigorous pre-commit reviews, it runs a ton of microbenchmarks & mutation testing. It writes better code than me.

2 4 185 13K 43

View Details

Cindy Sridharan @copyconstruct

19 hours ago

@pantasw2 Many projects for different reasons. SQLite for testing, llvm for layering + modular software design, scylladb for performant data structures, go std lib for canonical ways to handle/test http and other protocols, mpatterns in libraries like rust’s tokio, swift’s task/actors etc.

1 0 105 9K 175

View Details

Cindy Sridharan @copyconstruct

19 hours ago

@dcarter_js Might need to scrub some work specific stuff, but about 95% of it is just good general software engineering principles I’ve collected from the best sources I know (if I do say so myself). I’ll try to make it public over the weekend.

39 8 949 23K 582

View Details

Cindy Sridharan @copyconstruct

2 days ago

@treatyoma This admin is full of young people who deeply sympathize with this cancerous ideology, starting with JD Vance at the very top, who along with Elon let’s not forget, went to bat for the DOGE guy who tweeted “normalize Indian hate”. So yeah, expect nothing much to change.

1 0 8 619 1

View Details

Cindy Sridharan @copyconstruct

2 days ago

Cindy Sridharan @copyconstruct

6 years ago

Blogged: Post commit reviews Where I question if pre-commit mandatory code reviews are truly needed and what a post-commit review world would look like. Caveat: post-commit review doesn’t necessarily mean post-deploy review. bit.ly/2Oid8jG

7 22 118 0 48

18 5 131 23K 114

View Details

Cindy Sridharan @copyconstruct

4 days ago

@vaastav05 This means: huge volumes of raw, unsampled log data for agents. Humans just need accurate summaries, where they can have a high degree of trust in the agent’s analysis and deductions. Humans running queries and looking at charts is a thing of the past.

0 0 3 550 1

View Details

Cindy Sridharan @copyconstruct

4 days ago

@vaastav05 The post was already too long, so I couldn’t go much into the o11y shifts needed for “agentic operations”. But we do many tectonic shifts here. All existing o11y abstractions are outdated. The query interface is primarily agent native, with humans needing accurate summaries

2 0 2 489 0

View Details

Cindy Sridharan @copyconstruct

7 days ago

How to effectively debug with agents when you no longer read or understand the code In this post, I propose a debugging framework for agents that doesn’t rely on human knowledge of code to get to the right “root cause” The TL;DR are the screenshots. Link to post in the reply.

Cindy Sridharan @copyconstruct

a month ago

Writing a post about “how to effectively debug code when you no longer fully understand it”. This is real bug I had to fix in ~9K LOC (100% AI generated). None of the tactics I started with here helped “root cause” the issue, lol. Had to take a different approach to isolate it.

33 30 1K 116K 1K

38 113 2K 132K 4K

View Details

Cindy Sridharan @copyconstruct

4 days ago

@vaastav05 But much easier to build within a company, where they use shared infrastructure, shared libraries, shared security primitives etc. Harder for datadog to build a product that’s gonna need ssh access to your production hosts to gather more data when needed.

2 0 3 258 1

View Details

Cindy Sridharan @copyconstruct

4 days ago

@vaastav05 This is also why I’m sceptical of “AI ops” startups who claim they can do this, because it needs code level changes to emit all the right signals, and access to a user’s production environment to gather data not emitted by the code. That’s harder to build *generically*.

1 0 3 258 0

View Details

Cindy Sridharan @copyconstruct

4 days ago

@vaastav05 It totally should be able to back up the hypothesis it thinks is the root cause using real facts. When it extrapolates, it needs to establish provenance at every step. This is doable; the biggest challenge is getting all the data to an agent (I act as a meat proxy for this)