Join 43,000+ subscribers an receive weekly byte-sized summaries of system design theory, real world systems, post mortems and career advice every week!
Joined November 2024
- Tweets122
- Following44
- Followers23
- Likes233
ByteSizedDesign retweeted
Issue #88 of iOS Code Review is out 📬
📬 This week we are lucky to be sponsored by Runway, whose Triage tool pulls every app issue into one inbox!
Check out the edition!
- Xcode 27.2 beta swaps pbxproj for a JSON project file
- The iPhone Duo simulator is here in Xcode 27.1 beta
- Toolbars go vertical on Duo, plus how to backport them
- A launch arg that finds your slow ForEach
ByteSizedDesign retweeted
Issue #87 is out and it covers what changes for you:
April 2027 deadline to build with the iOS 27 SDK
Xcode agents now boot simulators and tap your app
A Swift 6.4 source break and its fix
ArrangementView for iPhone Duo
Five agents each re-reading a 4,000-token brief is 20,000 tokens before anyone does any work.
One agent reads it once.
Default to fewer, fatter agents until you can show the number that says the split pays. New edition below!
ByteSizedDesign retweeted
A new model (TANGO) predicts all 29 joint degrees of freedom directly, one network, no separate navigation and control layers.
It beat a baseline that gets to use LiDAR on top of cameras. TANGO only gets cameras. Collision rate: 9.9% vs 15.8%.
On a real Unitree G1, it succeeded 10 out of 15 cluttered navigation trials versus 6 out of 15 for the baseline, trained zero-shot with no real-world data at all.
The whole training set cost 211 GPU-hours. Decomposing the problem was making it worse, not simpler.
This and more in the latest edition!
Just spotted Byte-Sized Design at #6 on Substack's Rising in Technology chart. 🎉
Genuinely floored, this newsletter is nothing without the 43k+ engineers who read, share, and reply every week.
Thank you for trusting us with your inbox.
More deep dives and roundups coming. Onward.
Stripe replaced Envoy with their own Go proxy and halved both CPU and latency.
At their scale that's ~3,000 days of inter-service latency saved every day.
In this edition we explore why they left Envoy after 10 years, and the connection-count trap that pushed them:
Stripe just quietly killed Envoy (the proxy literally every infra team reaches for by default) and built its own from the ground up.
Their reasoning: at 1.6% of global GDP flowing through their pipes, "pretty reliable" isn't a tier they can afford to live in.
Breakdown coming this week!
ByteSizedDesign retweeted
OpenAI just shipped GPT-6 Astra and had to gate part of it, the model got good enough at finding zero-days on its own that it crossed their "critical" cybersecurity line.
If you're on a security team, that vetted defender access is worth requesting.
The 99.9% ARC-AGI-3 number going around is from a favorable harness though. Standard harness: 62.7%.
ByteSizedDesign retweeted
Models are learning to spend reasoning tokens only where a problem needs it: one paper cuts MATH-500 tokens 41% with zero supervised data.
Meanwhile 3 papers show memory-augmented LLMs getting worse with accurate, relevant retrieval.
15 papers, test-time compute to robots. Link Below!
Read the full edition read.bytesizeddesign.com/p/h…
It takes four vendors to summarize a phone call.
One transcribes. A regex redacts. An LLM summarizes. Compliance reviews all three.
Three retries per layer, three layers deep: 3³ = 27 requests from one failure.
New edition, together with @AssemblyAI
GitHub, Aug 17: down 7h47m.
One Istio sidecar hit its concurrency ceiling. The autoscaler watched app CPU and saw headroom.
Server side was fixed in 4h. A VS Code retry bug ran the other 5, at 100k rps.
Ended with a 403s.
#SRE #DistributedSystems
Most retry logic treats a 500 as "try again" and a 403 as "this will never work."
GitHub used that to end its 8-hour outage: they returned 403s on purpose, recruiting the clients' own error handling to do the load shedding.
How one sidecar became a retry storm 👇
read.bytesizeddesign.com/p/g…
ByteSizedDesign retweeted
Issue #84: SwiftUI's hidden orphan switch, toolbars get priorities, and Media Intelligence in iOS 27
ioscodereview.com/issues/iss…
GitHub's remediation list from Monday's outage has five items. Four are bug fixes. The fifth is "review retry limits and backoff behavior across gateways and clients."
That's the only structural one, and scoping it as a review undersells it.
Retry policy is a capacity decision that nobody owns. One team adds a retry to a flaky call. A client team adds retries because laptop networks are bad. Both correct in isolation. The sum is a load multiplier that only activates under stress.
The autoscaling misconfiguration gets fixed in an afternoon. The retry architecture is still there waiting for the next trigger.
ByteSizedDesign retweeted
New paper deep dive: Metis, a "memory foundation model"
Stop bolting memory onto LLMs with RAG and vector DBs. Build it into the weights.
Each transformer layer gets a small dynamic memory matrix. Remembering, updating, and forgetting happen inside the forward pass. Updates are gradient-free, all learned weights stay frozen, and the memory state is fixed-size, so cost doesn't grow with conversation length.
Does it work? With the original context deleted, Qwen3.5 scores ~0 on LoCoMo. Metis-27B (same frozen backbone + memory blocks) scores 26.7. And the memory state compresses to rank 64 with 99.9% of performance intact, which makes per-user memory look like a LoRA adapter.
The honest caveats: capacity tops out around a few hundred words per update, and stored junk bleeds into general reasoning (IFEval drops 22 pts).
Full breakdown in this month's State of AI, link below 👇
stateai.substack.com/p/metis…