Jarod Taylor (@jarodtaylor) profile picture

18 years in software. Leading squirrels every day. 10x builder of 20 projects at the same time only to abandon them at 80% to chase another squirrel.

Dallas, TX
Joined November 2008
Photo by Jarod Taylor (@jarodtaylor)Photo by Jarod Taylor (@jarodtaylor)Photo by Jarod Taylor (@jarodtaylor)Photo by Jarod Taylor (@jarodtaylor)Photo by Jarod Taylor (@jarodtaylor)Photo by Jarod Taylor (@jarodtaylor)Photo by Jarod Taylor (@jarodtaylor)Photo by Jarod Taylor (@jarodtaylor)Photo by Jarod Taylor (@jarodtaylor)Photo by Jarod Taylor (@jarodtaylor)
Jarod Taylor retweeted
Tip: Switch Claude Code's output style to Concise to get short, straight-to-the-point results. Enable it via /config → Output style, or add "outputStyle": "Concise" directly to your settings.json. You can still get the full details when you ask.
150
81
33
1,286
95,319
Jarod Taylor retweeted
Efton Chism right now
Sources: The #Bills are releasing safety Chauncey Gardner-Johnson. Gardner-Johnson, who has started 6 games for Buffalo, did not play on Sunday due to a 1-game suspension for a spitting incident in the Raiders game.
7
153
3
2,211
107,398
Jarod Taylor retweeted
Every developer needs an AI shed: An always-on machine running on their tailscale network where the majority of their herdr agents are running.
624
358
208
7,759
660,480
Jarod Taylor retweeted
To make your agent produce high-quality output, equip it with a way to check its work. Tests are part of that. Here are some of the ones I think are worth investing in. - Write end-to-end tests that simulate real user flows and act as the ground truth - Use property-based testing to specify what should never happen and generate thousands of test cases that attempt to break those constraints - If replacing a system, compare the old and new versions using randomly selected inputs - Require tests to run quickly and deterministically, which reinforces the agent to retry when the loop is slow or unreliable (the agent may learn to retry, not fix the problem) Example tests say what should happen; properties say what must not.
105
92
7
1,175
63,950
Jarod Taylor retweeted
We are making Markdown files work natively across Google Drive and Docs. You can now preview .md files in Drive, open them in Docs, edit them, comment on them, and collaborate without converting the file into a Google Doc!
937
1,342
952
21,018
1,446,983
Jarod Taylor retweeted
I’ve spoken to over 50 people 1:1 over the last several months to better understand how they code with AI. It’s an absolute mess. Here are the top 8 mistakes I see. 0:00 - Intro 1:19 - Mistake 1: Talking to 2026 models like they’re 2025 models 2:43 - Mistake 2: Devising funky metaphorical factories to orchestrate your agents 3:59 - Mistake 3: Treating context as sacred 6:02 - Mistake 4: Treating previous conversations as sacred 7:59 - Mistake 5: Having a todo list 10:00 - Mistake 6: Using funky tools 13:47 - Mistake 7: Stressing over code quality 24:27 - Mistake 8: Using branches and worktrees
172
130
38
1,939
338,001
If this isn't a motivation for you to leverage AI to build... I don't know what is.
that’s why he is successful
1
144
Using @every compound-engineering plugin across harnesses has been mind-blowingly good. @trevin and @kieranklaassen definitely dogfood their own product. Every update is 🤌 I've been using @herdrdev for multi-harness orchestration where Claude Fable/Opus is the Architect/Planner/Orchestrator and Codex GPT 5.6 is the executor. The cross-model collab is so f'n good.
1
82
Ya'll wanna go see a dead body?
The bad guys have evolved.
77
This makes me feel good about my ADHD half-baked brain dumps I give my agents.
One pattern I find useful for working with LLMs is a nice long ramble session. Sometimes the LLM needs more bits to understand what you're trying to achieve, but you're too lazy to type them. In these cases I like to lean back, switch to /voice and just ramble for like 10 minutes, total mess, anything goes, full stream of consciousness. Sometimes I declare it up top, something like "switching to speech recognition sorry for any typos...". Sometimes I turn it into a small interview of a few turns. But I find that the LLMs are somehow very good at reconstructing long incoherent rambles and often their echo of your own tangle of thoughts comes out quite a bit cleaner than what you started with. The result is that you improve the mind meld and have to correct things less from that point on.
1
64
Beginning July 20, Claude Fable 5 will be included in all Max and Team Premium plans, at 50% of limits. Pro and Team Standard users will continue to have access to Fable via usage credits, and will receive a one-time $100 credit. Demand for Fable has been challenging to predict, which is why we rolled it out to subscription plans in stages, extending access several times as we secured additional capacity.
63
What a time to be alive.
Cloudflare just opened a waitlist to charge AI agents a fraction of a penny to access any webpage, dataset, or API. Think about what that means. The internet was built for humans. Humans look at ads, ads pay the bills. But in a few years agents will outnumber us, and agents don't click ads. So the whole model breaks, and something new takes its place: agents with wallets, paying per request. The arbitrage is obvious once you see it. Own a dataset agents constantly need, put it behind a tiny paywall, and let millions of agents pay you fractions of a penny on repeat. Same thing happened to music. Piracy was the wild west until Spotify and licensing showed up and someone got very rich building the tollbooth. The agent internet is about to have its tollbooth moment. 12-18 months to be early.
1
53
Looks like @OpenAI added new reasoning options!
60
Jarod Taylor retweeted
(1) Today we're releasing Muse Spark 1.1 -- a strong agentic and coding model at a very low price. It's available through our new Meta Model API and in Meta AI.
5,313
3,537
3,602
45,168
23,720,147
Jarod Taylor retweeted
The computer is being reinvented in the agentic era: - The model is the new CPU. - The harness is the new OS. - Hallucinations are the new bugs. - The context window is the new RAM. - Skills are the new apps. - Markdown files are the new config. - Evals are the new QA. - Context is the new moat. - Permissions are the new firewall - Trust is the new bottleneck. - Prompt is the new programming language - Agent is the new software. Anything you dream of, you can build. This is the greatest time ever to be building with computers.
226
494
66
3,395
205,922
This might be one of the clearest public explanations from Anthropic on how models and effort work, but it's still a bit frustrating that they haven't added intelligent routing or strong defaults. They have the telemetry to know which tasks which model/effort combination succeeds on, but they still put the full decision burden on us (the user). They also repeatedly give the vague "fix your context first" suggestion with zero examples or patterns for what that actually looks like to reduce the need for the higher effort and stronger models. My biggest takeaway (most actionable): When output is wrong: 1. Maximize context, skills, CLAUDE.md, and scoping. 2. Still wrong? Did it not know enough? -> larger model Did it not try hard enough (skipped files, no verification, gave up early)? -> higher effort
75
Jarod Taylor retweeted
GPT-5.6 sol launches thursday! happy building
1,853
1,870
686
31,570
2,166,398