On 2026‑10‑23 I shared two LinkedIn posts that dive into the performance limits of AI‑assisted coding and Zig’s explicit allocator model. The first explains why a 10× boost in coding speed does not translate to a 10× boost in shipping speed, using Amdahl’s law. The second outlines Zig’s allocator abstractions and why they are a good fit for building high‑performance databases.
Faster code generation increases the load on every stage after it, so the review queue grows and the bottleneck shifts.
Allocator is just an interface
ArenaAllocator frees everything at once once you are done.
FixedBufferAllocator allocates from a buffer you provide, often on the stack, with no heap involved.
std.testing.allocator fails your test if it leaks, so you get leak detection.
When coding is 30 % of the end‑to‑end cycle and AI makes it 10× faster, the overall speedup is about 1.37×. If coding is 50 % of the cycle, the overall speedup rises to roughly 1.8×, illustrating the diminishing returns described by Amdahl’s law.
To truly ship faster, I should measure the entire “idea → production” path, locate waiting points, and apply AI to downstream stages such as review assistance, test generation, flaky‑test triage, and deployment safety checks.
Irrespective of whether software engineering is coming to an end or not, think of it as a roller coaster ride.
You are already on it, so have fun while it lasts.
Either it crashes or lands you safely.
You have to be optimistic about technology.
If AI is here to stay and you refuse to adapt, you will be left far behind.
Your best bet is to stay positive with tech
it has come down to this 🤦
Anthropic merged 30 K pull requests per day, and at that scale Git became a bottleneck, leading them to invent a custom VCS.
I decided not to switch to a new VCS unless my team experiences a similar massive PR throughput that makes Git a proven performance limit.
My guidance: address real performance bottlenecks first and avoid cargo‑cult adoption of complex tools you don’t need.
I shared this perspective in a LinkedIn comment on 2026‑10‑14, noting that for most teams the simple git push workflow is still sufficient.
On 2026‑10‑07 I posted two LinkedIn updates. The first was a short personal note about waiting for AGI and UBI. The second announced that GitHub recorded 7.38 billion commits in September and is re‑architecting its infrastructure based on five guiding principles.
GitHub saw 7.38 billion commits in September 2026.
Decision: GitHub is re‑architecting its infrastructure to scale further.
- Systems coordinate only when needed
I emphasized that these principles are “fundamentals” that outlive the current architecture and must be revisited whenever systems drift at scale.
Personal sentiment: “Others are waiting for AGI. I am waiting for UBI. I am done, bro :)”.
- Do as much as possible in parallel
- Compaction and GC to happen async
- Trim down hot paths
- Separate storage from compute
In early October 2026 I posted a LinkedIn update asserting that economics will always favour the machine, reflecting a deterministic view of automation's impact on markets.
economics will always favour the machine.