Marc Brooker
@MarcJBrooker · joined 12 Oct 2013
Distinguished engineer at AWS. AI, agents, databases, and serverless. Views are my own.
27 477Followers
749Following
3 450Posts total
375.5KViews on collected posts
Последние посты
Ищем новые посты…
@MarcJBrooker Not even self-review of what an LLm generates?
I often see so many awkward constructs and even things like naming that aren’t caught even after AI code review, that I don’t think I’ll ever ship anything without a self-review.
I’m still shipping a lot, so it’s not
18.7K views · 53 likes
· 0 reposts · 5 replies
Open on X →
I see this transition as inevitable. In some places, it has already happened. In others, it still sounds entirely unbelievable.
But it's going to happen.
20K views · 160 likes
· 0 reposts · 4 replies
Open on X →
Some level of understanding of and mechanical sympathy for the implementation layers will still be a valuable skill. But day-to-day interaction and understanding will be entirely at a layer above.
21.9K views · 97 likes
· 1 reposts · 2 replies
Open on X →
That's a harder one, and the solution is going to come as part of a much larger industry-scale raising of the level of abstraction developers work at from "in the code" to "in the specification" (or "in the requirements" or "in the feedback").
23.3K views · 110 likes
· 1 reposts · 2 replies
Open on X →
Code review's value as a tool to increase software quality are numbered, and the end may already be here (but unevenly distributed).
But what about code review's value as a learning tool?
26K views · 101 likes
· 2 reposts · 1 replies
Open on X →
What does the future look like?
Review tools, powered by LLMaaJ-style techniques, static analysis, and automated reasoning methods. Principled approaches to testing and validation (PBT, DST, model-guided fuzzing, etc). Correct-by-construction techniques.
47.6K views · 276 likes
· 4 reposts · 7 replies
Open on X →
The combination of these things will lead to (and in many cases already lead to) better outcomes than traditional code review, both for large and small issues. Soon, much better outcomes than code review.
28.1K views · 108 likes
· 0 reposts · 1 replies
Open on X →
Short-term, many teams do it for good reasons. Quality. Understanding. Compliance. Those reasons exist, but we should be (and are) working on ways to make them go away.
48.8K views · 176 likes
· 0 reposts · 4 replies
Open on X →
The "mixed mode" where humans look at some code reviews, assisted by review tools, is also valuable today. But probably even more transient.
The idea that humans will reliably look through code to find the increasingly rare issues that automated tools miss seems like a fantasy.
42K views · 210 likes
· 6 reposts · 3 replies
Open on X →
If somebody at Apple wants to become my new best friend, all they need to do is backport the MacOS 27 endpoint security feature set to MacOS 26.
7.8K views · 27 likes
· 1 reposts · 3 replies
Open on X →
Only read transactions on schemas that specifically use foreign key constraints, to be clear.
But if you're using FKCs and coordinating on reads, why not be pessimistic?
3.6K views · 7 likes
· 1 reposts · 2 replies
Open on X →
@MarcJBrooker I wonder, at this point, whether PCC would be a better fit, given that even read transactions can now abort at commit time.
3.8K views · 0 likes
· 0 reposts · 0 replies
Open on X →
Did you know that Aurora DSQL now supports foreign key constraints?
The team's done a great job here: FKCs that scale well, especially for your read-heavy workloads.
Check it out here: https://t.co/LkGszVsKFu
Quick thread on how FKCs in DSQL work... https://t.co/kTK34XAqMI
9.5K views · 88 likes
· 15 reposts · 5 replies
Open on X →
@MarcJBrooker out of curiosity, have you observed fsync failures in production before?
979 views · 7 likes
· 0 reposts · 2 replies
Open on X →
It requires the assumption that fsync() failures are uncorrelated. They may be correlated between machines (e.g. by temperature or power quality), but highly unlikely to be correlated between AZs and regions.
1.2K views · 9 likes
· 0 reposts · 3 replies
Open on X →
This is converting a complex/partial/gray failure into a simple "fail stop" failure. The ability to do this is a great system simplifier.
1.2K views · 14 likes
· 0 reposts · 1 replies
Open on X →
Love this topic, because it's a perfect illustration of why so much of the "single system is simpler" discourse is wrong (because it mostly ignores the hard problems that introduces).
1.3K views · 19 likes
· 0 reposts · 0 replies
Open on X →
Our paper doesn't go into much depth on this, but will give you the context you need: https://t.co/uSRWI4xVAa
1.4K views · 17 likes
· 0 reposts · 1 replies
Open on X →
Case 1: sync failure on commit (in Journal). Journal has multiple sync replicas and multiple async replicas. An fsync failure in a sync replica will cause it to be removed from the quorum, and replaced by an async replica.
1.9K views · 21 likes
· 0 reposts · 1 replies
Open on X →
Edge cases are handled by the reconfiguration protocol, which makes sure there is a consistent monotonic view of the data. This protocol needs to exist anyway (to get multi-AZ or multi-region fault-tolerant durability).
1.6K views · 13 likes
· 0 reposts · 1 replies
Open on X →
Case 2: sync failure on apply (in Storage). Storage isn't responsible for durability, and uses flash (essentially) as an expanded/cheaper memory. Recovery is from Journal/S3.
So could safely ignore an fsync failure (but doesn't because it's a symptom of deeper issues).
1.3K views · 12 likes
· 0 reposts · 1 replies
Open on X →
Handling edge-case failures like this is way simpler in distributed storage databases (e.g. Aurora) and distributed databases (e.g. Aurora DSQL) than in single-system databases. It's one of many cases where distribution simplifies system properties.
Here's how it works in DSQL:
27.1K views · 234 likes
· 16 reposts · 5 replies
Open on X →
Does anyone know good blogs around how databases handles fsync failures?
26.8K views · 27 likes
· 2 reposts · 7 replies
Open on X →
Absolutely beautiful part of the world.
Just don't mess with the hippos.
9.4K views · 22 likes
· 0 reposts · 1 replies
Open on X →
На фоне аккаунтов своего размера
24 постов за последние 90 дней рядом с диапазоном 10K–100K подписчиков. показывается широко, но откликается мало кто.
Медианные просмотры9 480этот аккаунт924медиана для 10K–100K
Охват, %34.50%этот аккаунт3.62%медиана для 10K–100K
Вовлечённость, %0.57%этот аккаунт1.52%медиана для 10K–100K
| Показатель | Этот аккаунт | Медиана для 10K–100K | Отношение |
|---|---|---|---|
| Медианные просмотры на пост | 9 480 | 924 | 10.3× |
| Охват (просмотры ÷ подписчики) | 34.50% | 3.62% | 9.53× |
| Вовлечённость | 0.57% | 1.52% | 0.37× |
Другие в этом диапазоне → Сравнить с другим аккаунтом → Как считаются эти ориентиры →
Growth & engagement
How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.
Views per post
97918 Aug
9.5K31 Aug
3.8K
3.6K1 Sep
7.8K2 Sep
42K18 Sep
48.8K
28.1K
47.6K
26K
23.3K
21.9K
20K
18.7K
Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
0.92%18 Aug
1.16%31 Aug
0.03%
0.28%1 Sep
0.41%2 Sep
0.52%18 Sep
0.37%
0.39%
0.61%
0.40%
0.49%
0.46%
0.82%
0.31%
Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.
What the audience does
Likes70.9%1 808 in total
Reposts1.9%49 in total
Replies2.4%62 in total
Quotes0.5%13 in total
Bookmarks24.3%619 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.
Похожие аккаунты
Anthony DiMichele ✓
@Coach_DiMike
55K followers
· 3.4K posts
Frontier Airlines
@FrontierCare
55K followers
· 213.4K posts
Kabiran Nabirah
@kabiran_nabirah
55K followers
· 196.7K posts
tam lu
@tml096
55K followers
· 106 posts
911 Emergencias Pue ✓
@911Puebla
55K followers
· 79.2K posts
Kiplinger
@Kiplinger
55K followers
· 84.1K posts
Chad P. Bown ✓
@ChadBown
54.9K followers
· 19.1K posts
DIG RANGE BASTI
@digbasti
54.9K followers
· 16.2K posts