Michael Francis
@edude03
Followers
484
Following
39K
Media
282
Statuses
8K
Software developer who loves infra. Recovering Kube addict, dabbles in ML so the kids think I'm hip.
Toronto, Ontario
Joined December 2008
A major red flag for me is when a person making architectural suggestions or decisions for a non-FAANG begins, with literally no data informing the point, that you should put a cache (Redis, etc.) in front of your transactional database (PostgreSQL, etc.) Even if they insist it’s
There’s no such thing as “scalable apps” only “scalable databases”. Don’t allow a DBMS to dictate that: - your app “shouldn’t have hot keys” in your workload, or - any isolation less than strict serializability is ACID. Insist, per Jim Gray, that OLTP is “correct and fast”.
44
49
899
In fact redis is just a convenient way to do the wrong thing 90% of the time
1
0
0
Why would I trust someone who royally fd micro service architecture AND used Reddit’s as a cache to improve the situation in any way
In a distributed environment, one of your cache layers (e.g., Redis) fails. As a result, all requests hit the database, causing a cache stampede, where multiple microservices simultaneously query the database for the same data, leading to a bottleneck. How do you design a
0
0
0
no one prepared me for the fact that when you have a 9-5 you can't like. go to the dentist or the doctor or anything anymore. because you work. and when you don't work they're closed.
2K
15K
222K
just received a breakup text that started with Thought for 7s
174
150
10K
Derive is live on HyperEVM. HYPE holders can now use their tokens as collateral to trade options and perps on Derive. TVL stays on @HyperliquidX while users generate yield on Derive. HYPE becomes a yield-bearing collateral. Try it out at https://t.co/a1orxMLYqB Hyperliquid.
33
56
470
Antigravity seems like it'll be awesome once there is capacity to run it (it hasn't completed a task since I installed it)
0
0
2
kimi via openrouter via moonshots api is absolutely chugging, I thought y'all used claude and codex - usually see double the tps
0
0
1
Where’d we land on dense vs MoE models? Since mixtral 8x7b seems like we go back and forth on that
0
0
0
Who makes up these kind of debates when from scratch exists?
Most Docker debates eventually reach this question: slim or alpine, which base image should you use? Alpine is very small and looks attractive for cutting image size. But it uses musl instead of glibc, so some apps behave differently, some libraries don’t work as expected, and
0
0
0
Adding on to this - the biggest unlocks for me learning rust was 1) the builder pattern - understanding what “consuming” Self meant taught me mutability and 2) GADTs “enums” as a type that can contain a thing combined with match to unpack the thing. These two things are like 80%
For all the people who asks me on steps to learn rust. This is the best and most effective way to learn rust: Build something. It's that simple. Just build. For each problem you get, do some research, ask AI about input(don't let it code for you). For each problem you get will
0
0
1
Most junior backend devs think rust = fast then make blocking calls to their database that's also somehow in a different region
most junior backend devs think performance = faster code it’s not. performance is architecture caching, async I/O, efficient DB queries, rate limits, batching, pagination, logging you don’t optimize a function you optimize a flow
0
0
0
Where's my CNC nerds? Should I buy a Nestworks C500 or am I giving in to hype?
1
0
0
There is this show on YouTube called sorted where guys raced their purpose built race cars in a bunch of challenges. Everyone spent about 150k total, but most guys bought a 20/30k beater and spent the rest on upgrades, one guy bought an entry level Porsche. Not only did the
I tried to price out the actual tinybox v2 server components (minus GPUs and chassis), and I came up with $15k. The 5u 31" deep chassis is kinda hard to find, but I doubt it'd be more than like $1k So, $16k for the base server Add in 8x5090s @ 3k, $40k in components, $10k to
5
1
115
“My solution is better because you’re too dumb to know how to use other solutions”
"It's OK if we use an SPA because the cold start cost will only be paid one time" Ah cool, so your users will just never: a) open in a new tab b) have to refresh because 100% of state is synced via websocket c) hit a version mismatch d) open multiple windows side by side e) ...
0
0
0
Run gpt-oss-20b on openrouter get 32/100 on benchmark. Run gpt-oss-20b on vllm with h200s get 83/100 on benchmark. What are these providers doing? Deepinfra terrible results
53
17
585
Gossip and CRDTs are such a game changer for eventually consistent distributed systems
Forget Raft & Consul. https://t.co/9CHVCRhMh8’s Corrosion uses CRDTs + gossip + SQLite for global state sync—optimized for speed, not consensus. https://t.co/u7mXqBwXBa
0
0
1