
Building things on the web. Currently @HubSpot, formerly @fireworksAI_HQ, @wayfair, @rpi alum.
Boston, MA
Joined February 2010
- Tweets18.9K
- Following1.3K
- Followers1.1K
- Likes63.1K
increasingly surprised by just how little apple seems to care about apps in the app store
Tried searching for some apps like feed readers and video players the other day and it’s 90% slop now
Apple nominated these 4 as "gems that debuted in 2026 really stood out to us"
- a camera dictionary where even the screenshots are AI slop
- the most basic pencil note taker
- the most generic countdown widgets app (ok I'm biased)
- a planner with 5 UI styles that breaks half the Human Interface Guidelines
how does the App Store pick these?
26
Posted a few small updates I've been experimenting with and adopting in my own agentic workflows: matthamlin.me/2026/november/…
TL;DR: Orchestrator threads are powerful, and I recommend them currently!
1
40
Why should projects announce an upcoming release like this at all, as opposed to just dropping the fix on the pre-determined schedule?
It’s not like i can prepare to upgrade node right now😂
Node.js has been doing security pre-announcements for years. This isn’t something they started a year or two ago. It’s a deliberate, responsible decision.
Publishing patches whenever they’re ready, without advance notice, ignores the schedules of all the developers who need to respond.
Also, I actually trust projects that are actively working on security these days more. It shows they’re improving their workflows and actively making their projects more secure.
1
108
i used to think compaction was solved and good enough that i didn’t have to worry about it, but maybe this is a far better strategy
Compaction is a mistake.
When an agent's context fills up, tools like Codex and Claude Code compact it:
1. The model prefills the entire session again
2. It writes a summary of everything so far
3. A new session starts from that summary
I measured this on a real ~400k-token session with Opus 5.5:
→ 38 s and $1.99 when the prompt cache has expired
→ 29 s and $0.17 when it hasn't
And the summary is lossy. The exact error message, the file path, the one number you needed is often gone.
So I do something new for Mainplane
Context retrieval works the other way around. Mainplane keeps the full record of every session as a file. When the context fills up, a new session starts right away with one instruction: "You are continuing session X. Retrieve key context from it as you go."
The agent searches the old session and copies the exact lines it needs, when it needs them.
Same session, same model: 15 s and $0.10 to get back to work. It does not depend on a cache hit, and nothing is paraphrased.
This is the bitter lesson applied to agent memory. Don't hand-design what the model should remember. Give it the full record and good tools, and let it decide.
Mainplane is open source (MIT). Link in the first comment.
2
42
i’ve been referring to this as context poisoning
still haven’t found a good solution to it yet though
Current agent pet peeve is how tainted the codebase gets with artifacts from the dialogue.
> tell agent to do X
> agent does X, but also does Y
> tell agent I don't want Y
> agent removes Y, but leaves comments everywhere like "X does x; don't worry, we don't do Y"
43
my biggest gripe with @t3dotcodes right now is that i can’t use the built in pull request view for prs associated with the thread but in a subdirectory (or anywhere else not in the current working directory)
@theo think you can include this in your upcoming updates? 👀
1
40
Really tempted to try and reverse engineer the Arc browser and then offer a $1/year subscription to cover maintenance costs on it
1
57
Would want to include tab syncing across devices / support for mobile apps (which is one of the features that most Arc replacements do not offer), and optional AI features powered by your own subscriptions
1
40
Ok @executorsh just fixes this entirely: executor.sh/
why is it so difficult to setup MCP servers with ChatGPT, it feels like theres many ways to configure it but they all never work as expected
1
1
130
why is it so difficult to setup MCP servers with ChatGPT, it feels like theres many ways to configure it but they all never work as expected
i still have yet to see a clear answer to why the browser company dropped arc and built a worse version of it
1
1
55
i was wondering why the embargo for iPhone duo reviews was after the preorder date, and i think it’s because the apps are just not even anywhere close to ready for the phone (and maybe apple figured that’d be the case when setting the embargo date)
2
208



















