Erel

eng @greptile @gtalumni

[Erel @ozegaard]

@greptile computer use

[Erel @ozegaard]

Excited to be a part of this team. If working on difficult problems in Engineering or Research interests you, reach out!

[Greptile @greptile]

Greptile has raised a $50M Series B led by Menlo Ventures. The round values the company at $750M.

Greptile reviews and tests more than 1% of all PRs opened on GitHub every day. Thousands of engineering teams including Nvidia, Coinbase, Decagon, and Vanta use Greptile as their independent and central validation layer.

Fewer than 50 people, we are a small, talent-dense team of engineers and researchers in a quiet corner of San Francisco. If you’re interested in joining us, we’d love to hear from you.

[Erel @ozegaard]

the sf housing market has gone too far

[Erel @ozegaard]

greptile

[Felix Fong @_fuccii]

greptile

[Erel @ozegaard]

@madsf88 greptile

[Erel @ozegaard]

bouta show this nerd school what real football looks like @Stanford @GeorgiaTech

[Greptile @greptile]

Today we’re releasing new Greptile review tiers: Plus and Apex.

Higher tiers have more tools and subagents, can run tasks in parallel, better understand the PR, and catch more bugs.

- Base: Thorough code review with full codebase context.
- Plus: A deeper review for PRs that need extra scrutiny.
- Apex: Our most intensive review for large, complex PRs.

[Erel @ozegaard]

The experience of using Claude code or Codex is analogous to eating blue cheese

[Erel @ozegaard]

Life is an honest caveat

[Erel @ozegaard]

@ChipsAhoy can you make this please

[Erel @ozegaard]

child restrictions on an iPad is just sandboxing your agent

[Erel @ozegaard]

a little side project

Title: Kernel Queues vs. In-Process Schedulers: A Performance Analysis https://t.co/htKKYHZGbM

[Erel @ozegaard]

@muhammadyii For local data based MCPs, what I said still applies. Otherwise I agree with you.

[Erel @ozegaard]

GPT 3.5 turbo had 16k context, and 2.25 years later gemini-2.5 has 1M context.

If we see the same rate of growth, in fall of 2027 we should have an LLM that can fit the entirety of the Linux code base in its context, no RAG, no MCP, etc

[Erel @ozegaard]

@muhammadyii Not entirely, MCPs allow interaction with data sources, but what if you could just ingest the entire data source? Not feasible for large sources, but things under a gigabyte could be possible in the future

[Erel @ozegaard]

@r1sheeb assuming exponential is bullish, so I’m just gonna say linearly for now

[Erel @ozegaard]

@alexandr_wang @Meta @natfriedman what in the 1992 dream team is this

[Erel @ozegaard]

@r1sheeb Didn’t that come out in 2020

[Erel @ozegaard]

I hate when a coding agent is like

“You’re so right!“

and then it seg faults again