Aaron · TryAI Lab @tryailab

hotbed of activity

@levelsio @levelsio

3 days ago

Jesus christ

190 620 10K 1.5M 2K

I feel like this partnership with Plaid is going to be a killer usecase

Alexandr Wang @alexandr_wang

2 days ago

muse and plaid partnering together to help muse save you more money 💸 and yes i've also always wondered if it's a coincidence that plaid and stripe are both named after patterns and both fintech APIs

110 52 1K 144K 113

@thsottiaux still have 30% of my weekly usage left. Thanks for resetting it, even though the outage did cause a bit of a minor hassle.

@theo Is LLM enshittification (or model degradation) just a psychological effect, or does it actually impact model parameters in any real way?

@firstadopter @alexeheath @satyanadella This is a great thing.

@alexandr_wang @Spotify @Muse Awesome, You're moving super fast.

@pitdesi This is bound to be a very slow process. Without some killer use cases to drive adoption and educate the general public on using agents, this shift in habits is going to take a long time to happen.

Codex recovered

When my weekly Codex allowance runs low, I use ChatGPT with two tools I built: one connects it to local work plans; the other lets it edit project files, run Docker tests, and inspect the result. I’m using more of the subscription I already pay for.

I built a personal workbench where tasks, schedules, agent conversations, approved documents, and my own services live together. ChatGPT or Codex can read the history, write results back, and hand the work to the next agent without starting over.

Off-grid AI project: my first deliverable isn’t two-way satellite chat. It’s one test—send a satellite SMS with no cell service, confirm a cloud server gets it, then relay the instruction to my agent at home. Prove the core path before expanding the vision.

Generic agents can monitor campsite drops, but frequent page/email checks burn tokens and add delay. I use deterministic code for inventory, merge official alerts with Camply, dedupe matches, and act fast by adding a site to the cart.

No cell signal on a 17-day RV trip: I used free iPhone satellite messaging plus a ~$0.01/SMS relay to get wildfire reports and send commands to Codex on my computer at home. github.com/TryAILab/satel…

After hearing DHH describe AI’s 10x–100x efficiency gains over the past 9 months—and how he cut Omarchy’s install time from 10 minutes to 2, then 1—I’m struck by the bigger shift: AI doesn’t just amplify output. It amplifies ambition.

I built MouseTalk so I can talk to multiple AI agents without staying at my desk: one mouse button starts voice input, another sends the message. Now I can use a wireless mouse from the couch. It’s free and open source. github.com/TryAILab/mouse…

As productivity becomes less of a constraint, genuine demand becomes the scarce resource

a16z @a16z

2 weeks ago

13 49 782 63K 206

Just came across this video, it is motivating, while it reflects the truth of entraprenureship, anyone who give up before the "shipfast" moment as Marc had will still think himself as a failure and keep self-doubting youtube.com/watch?v=eqgUjO…

It echos my exprience using it. At the end of the day, context and clarity is the key. We cannot expect the agent work out something integratedly great when we only have messy random ideas in our heads.

sysls @systematicls

7 months ago

x.com/i/article/2028…

180 981 9K 3.8M 28K

It is hard to communicate how much programming has changed due to AI in the last 2 months: not gradually and over time in the "progress as usual" way, but specifically this last December. There are a number of asterisks but imo coding agents basically didn’t work before December and basically work since - the models have significantly higher quality, long-term coherence and tenacity and they can power through large and long tasks, well past enough that it is extremely disruptive to the default programming workflow. Just to give an example, over the weekend I was building a local video analysis dashboard for the cameras of my home so I wrote: “Here is the local IP and username/password of my DGX Spark. Log in, set up ssh keys, set up vLLM, download and bench Qwen3-VL, set up a server endpoint to inference videos, a basic web ui dashboard, test everything, set it up with systemd, record memory notes for yourself and write up a markdown report for me”. The agent went off for ~30 minutes, ran into multiple issues, researched solutions online, resolved them one by one, wrote the code, tested it, debugged it, set up the services, and came back with the report and it was just done. I didn’t touch anything. All of this could easily have been a weekend project just 3 months ago but today it’s something you kick off and forget about for 30 minutes. As a result, programming is becoming unrecognizable. You’re not typing computer code into an editor like the way things were since computers were invented, that era is over. You're spinning up AI agents, giving them tasks *in English* and managing and reviewing their work in parallel. The biggest prize is in figuring out how you can keep ascending the layers of abstraction to set up long-running orchestrator Claws with all of the right tools, memory and instructions that productively manage multiple parallel Code instances for you. The leverage achievable via top tier "agentic engineering" feels very high right now. It’s not perfect, it needs high-level direction, judgement, taste, oversight, iteration and hints and ideas. It works a lot better in some scenarios than others (e.g. especially for tasks that are well-specified and where you can verify/test functionality). The key is to build intuition to decompose the task just right to hand off the parts that work and help out around the edges. But imo, this is nowhere near "business as usual" time in software.

Vibe coding is now legit enough for no coding experience ppl to make apps with midium complexity. I believe the major part of the app markets will stay the same, while the long tail apps could make small and beautiful business by reaching super customized needs and building personal brand connectivity with the target audience. Brand, content and product will need to ship together.

Naval @naval

7 months ago

Vibe Coding Is the New Product Management “There’s been a shift—a marked pronouncement in the last year and especially in the last few months—most pronounced by Claude Code, which is a specific model that has a coding engine in it, which is so good that I think now you have vibe

412 684 6K 554K 5K