another nugget of wisdom to add:
say "idk" more often and ask more questions
notice how enraged u feel when the LLM claims it knows something it doesn't actually know
agreeable people (who don't pushback or admit when theyr wrong) are a nightmare to work with if ur innovating
Think of yourself as an LLM.
Every social interaction, every meeting, burns your tokens.
Unless someone is a paid subscriber to your attention, you are under no obligation to answer low-quality prompts.
Think of yourself as an LLM.
Every social interaction, every meeting, burns your tokens.
Unless someone is a paid subscriber to your attention, you are under no obligation to answer low-quality prompts.
there's some truth to this but something tells me this is just a big PR move to further promote 'Mythos' for a better IPO, just to cover the ever-increasing costs of training new models...
BREAKING: Anthropic has urged for a global pause in AI development as artificial-intelligence models are nearing capability to improve without human intervention, per WSJ
No, @greptile + /self-heal for code reviews
periodic /improve-codebase-architecture + /desloppify to keep architecture clean and avoid technical debt build up
and browser QA with https://t.co/Pym8ClkpBZ + https://t.co/5nJQVOYLdD (magical if using parallel worktrees)
Sam Altman said AI budgeting has recently become a "huge issue" for some companies, something that "never came up" earlier this year. https://t.co/P2zODBNmDp
The better your observability is (what ur coding agent can see - logs, traces, latency and so on), the more you can afford to decrease the 'thinking' level on the model
tighten that feedback loop and you'll create outsized impact
not enough people talking about this
I wish Slack was:
- Agent-first
- Beautiful to use
- Integrated with agents natively so your Hermes or OpenClaw lives inside it
- Huddles worked seamlessly and were fun
- Built for teams of 1-3, not just teams of 300
- Truly a second brain similar to Obsidian
- Searchable without wanting to throw your laptop
- Designed around async, not constant interruption
- Voice first for mobile
- A place where I could see who's working on what right now without asking anyone
- Smart enough to know the difference between "I need you right now" and "whenever you get to this"
- A workspace where my agent could tap someone else's agent on the shoulder and coordinate without involving either human
- Designed so the new hire on day 1 has the same context as the person who's been there 3 years
-Something that felt like walking into a room of people building, not walking into a room of people typing
- A place where decisions are first-class objectsξ»ξΉξ
- Able to auto generate SOPs, skills, agents etc from conversation history
- Something that rewards deep work instead of punishing it with 47 unread notifications
You can just bs things.
Every month I have to mute performative accounts like this, who haven't built anything to save their lives.
Filtering reality from the noise is a skill in itself.
This is a good heuristic tho.
If they haven't built anything -> insta mute
Agentic Engineering has completely changed it's shape in the past few weeks if you want to get ahead here are some concepts you absolutely need to know for your own scaffolding
1. Continual Traces β Traditional logging has been rendered obsolete by these god tier models and harnesses. so instead you use continual traces logs that build upon themselves as the context grows
2. Agent Security Stack β Trust provision provenience and containment by design.
3. Stochastic Hierarchical Iterative Topology β Adaptive multi-level exploration of solution spaces. Using tree traces and the inherent laws of math to direct your agent harness for long running tasks
4. Relational Context Retrieval β Retrieve relationships, not documents, using relational dbs to constantly improve and connect your growing context
5. Filter Using Constraint Context: Current harnesses will have you believing that semantic search has been rendered obsolete but this technique allows you to improve your harness by leveraging the inherent semantic understanding of LLMs to efficently search relevant context.
DM me for the complete doc to 100x your agentic capabilities
@shaunmmaguire Nothing new. Amp has been doing this for a very long time, but curious how it compares.
It's an interesting problem. They're probably using a generally strong but more affordable model to route to the right model based on their own datasets
@housecor Yeh I've been thinking about this for a few months now, but basically everything would need to be heavily feature flagged, canary release almost all deployments and auto-fix High-confidence bugs as they arrive.
A few things to be careful of tho (DB migrations and so on)
PR 1000 landing today.
Code Factory is fully functioning.
Agents write/review/land 100% of our code now.
Don't get me wrong - there's a *lot* of work that I do crafting PRDs, steering conversations, etc, but the majority of the work is done by agents now.