⚡️ Trevor Hartman

5.3K posts

⚡️ Trevor Hartman banner
⚡️ Trevor Hartman

⚡️ Trevor Hartman

@devth

The faster I go, the faster I go ╲︿_/︺╲▁︹_/\_︿/﹀▔\⁄﹀\╱﹀

Montana Sumali Mart 2008
1.5K Sinusundan450 Mga Tagasunod
Naka-pin na Tweet
⚡️ Trevor Hartman
⚡️ Trevor Hartman@devth·
It turns out my MacBook fits perfectly in my office fridge 🤔
⚡️ Trevor Hartman tweet media
English
1
0
9
0
⚡️ Trevor Hartman
@ZubyMusic I’m also 1 of 5 and have 5 of my own. My goal was 5 by 35 and my 5th was born when I was 35! Best decision and infinitely more meaningful than any professional goals or achievements.
English
0
0
2
63
⚡️ Trevor Hartman nag-retweet
Dillon Mulroy
Dillon Mulroy@dillon_mulroy·
the most annoying behavior of 5.5 is that it regularly starts editing files with python instead of its edit tool
Dillon Mulroy tweet media
English
46
2
435
47.1K
⚡️ Trevor Hartman
this may continue for a few replies over a day or week, and typically result in consensus, alignment, and all engineers aware of the context and any decisions made. all without slack-DM insanity or zoomcall hell.
English
1
0
0
20
⚡️ Trevor Hartman
i miss email. in 2011, people on the engineering team would email a thoughtful question or proposal to all engineers, and within a few hours or the next day receive a very thoughtful reply.
English
1
0
0
26
NIK
NIK@ns123abc·
I’m almost done writing day 3 of Musk v. Altman Every exhibit traced Every quote sourced Every claim cross-checked Mainstream news doesn’t go this deep Truth 💣 dropping soon…
English
61
36
1.1K
49.7K
⚡️ Trevor Hartman
inspires confidence when opie47 gets stuck in its own wellackshualllyy loop
⚡️ Trevor Hartman tweet media
English
1
0
0
14
José Valim
José Valim@josevalim·
At @ElixirConfEU, @chris_mccord and I set up two races between myself and a coding agent. One was to add a new feature to DBConnection (max_lifetime) and another was to investigate a potential type system regression (links below). For DBConnection, I gave the documentation of how the option would behave. Not only was I faster, the coding agent solution had 4x more LOC than mine (it changed files and data structures that were not necessary for the feature). For the type system, the coding agent could not fix the problem at all. It did recognise it was a type system regression, and it did provide an initial fix in roughly the same time as I did, but the coding agent fix was wrong. After I fed it additional tests (from my own solution), it failed to come up with a fix. I use coding agents daily, I often ship their code as is, but they can still slow me down, they are nowhere close to fixing all problems I tackle in a given week, and are often not up to the standards I expect from my own software.
English
10
31
304
11.6K
⚡️ Trevor Hartman nag-retweet
José Valim
José Valim@josevalim·
At this point, this is just irresponsible. Yes, coding agents are leading to an increase of software production, but we are not seeing a similar push or increase in software quality. If Anthropic focuses on safety and it believes software engineering is going away, then it needs to be doing much more to improve how we design, build, test, and maintain software (aka software engineering). Increasing the production of unreliable, poorly designed, and unverified software directly undermines safety. Claude Code is claimed to be "fully written by AI". In the last two months, it took three separate postmortem-worthy failures and user complaints to surface what their own testing missed. Yesterday users were being over billed by hundreds of dollars. Software engineering isn't ready to go away and there is not enough progress to argue that case. I am certain Anthropic would argue that AI progress in other domains is strongly dependent on having proper safeguards in place. I can't wrap my head around the cognitive dissonance when it comes to software. PS: Mythos (may) improve software security, but that is only a subset of safety.
AI Edge@aiedge_

Anthropic CEO (Dario Amodei): "Coding is going away first, then all of software engineering." What do you think about this?

English
68
155
1.2K
141.5K
⚡️ Trevor Hartman nag-retweet
Gergely Orosz
Gergely Orosz@GergelyOrosz·
I am increasingly bullish on open source harnesses (like OpenCode) not because they will be better than SOTA closed harnesses, but because they will never pull shady stuff like what Claude Code and other closed harnesses can and do, regularly. Every incident like this a reminder
Om Patel@om_patel5

THIS GUY LOST $200 IN ONE DAY BECAUSE THE STRING "HERMES.md" WAS IN HIS GIT COMMITS HERMES.md is a real convention used in AI agent projects. it's a system prompt specification file. not some obscure edge case he's on claude max 20x at $200 a month. yesterday claude code hit him with "you're out of extra usage" out of nowhere his dashboard showed 13% weekly usage. 0% current session. 86% of his plan was sitting there untouched but $200.98 in extra usage already burned through what should have been covered by his subscription he tried logout & login, different models, fresh installs and nothing worked anthropic support sent the ai bot (four rounds of the same scripted response). eventually they just gave up on him so he started binary searching repos and commits manually on his own time until he found the trigger the string "HERMES.md" in a recent git commit message uppercase, with the .md extension, anywhere in your commit history that's it claude code includes recent commits in its system prompt and something server side flags HERMES.md and quietly routes you off your max plan onto API rate billing > AGENTS.md? fine > README.md? fine > HERMES without .md? fine > lowercase hermes.md? fine > uppercase HERMES.md? you're getting charged API rates he reported it. anthropic support acknowledged the bug three times, called it an "authentication routing issue", thanked him for finding it then refused to refund the $200 so the man pays $200 a month for max, lost another $200 to a billing bug they confirmed, did anthropic's QA work for free on his weekend, and got a "thank you for your patience" in return check your commit history before claude code quietly drains your account too

English
90
155
2.2K
220.9K
⚡️ Trevor Hartman
⚡️ Trevor Hartman@devth·
Any harness that’s coupled to a single AI vendor is lame and fundamentally brain dead.
Mark Kretschmann@mark_k

Grok Build is very close now. Release is expected next week unless something major causes a delay. Grok Build is @xAI’s answer to Claude Code and Codex. It consists of a suite of specialized coding models, variants of Grok 4.3, along with full tooling in the form of a web app and CLI. Grok Build is very important for both xAI and its users. Strong alternatives to the big two in AI coding are sorely needed.

English
0
0
1
23
⚡️ Trevor Hartman
⚡️ Trevor Hartman@devth·
@_ashleypeacock Would love it if billing could be baked into my monthly Cloudflare bill instead of having to load credits and pay fees!
English
1
0
0
110
Ashley Peacock
Ashley Peacock@_ashleypeacock·
A single line of code to access all major LLMs. No provider API keys. That’s how simple Cloudflare’s AI Gateway is - billing handled at cost by Cloudflare for any proxied model!
Ashley Peacock tweet media
English
13
7
181
13.2K
⚡️ Trevor Hartman nag-retweet
Rep. Anna Paulina Luna
Rep. Anna Paulina Luna@RepLuna·
If members of Congress are caught insider trading, it’s a small, couple-hundred-dollar fine. If you’re caught doing it in the military, like in MSgt Dykes’s case, you could be facing up to 50 years in prison. Anyone who cannot see that members of Congress are insider trading is either low IQ or willfully ignorant. Let’s take Nancy Pelosi, for example, it is statistically not possible for her to make a 17,000% return and outperform Warren Buffett without insider information. 🤡
English
1.5K
10.5K
37.8K
370.5K
Ashley Peacock
Ashley Peacock@_ashleypeacock·
Cloudflare Workflows steps now support ReadableStream as a return type This makes it trivial to handle larger outputs, such as reads from R2, that would exceed the 1MiB step limit
Ashley Peacock tweet media
English
4
3
117
6.9K