This means that, in practice, it functions more like an agent than a streaming text model. That in turn breaks trajectory analysis which not only makes distillation harder, it also makes it much harder to improve your own harnesses and agents built on top of the model.
Foam | Staff Founding Eng | In-Person, San Francisco | Full-time (M-F 9am-7pm) | $300-$400k/yr + equity
Foam is building a Better Sentry on an observability stack purpose-built for AI. We automatically identify and root-cause breakages across systems with zero prompting and zero configuration then DM the engineer responsible, not the git-blamed engineer.
We're looking for a hands on keyboard Staff Founding Engineer who's been unshackled by modern coding agents.
Here's a few of the problems we'd love your help on:
1. Reducing our onboarding time from hours to seconds
2. Scaling our telemetry engine into the petabytes
3. Collapsing our latency from minutes to seconds
4. Beating our intern at chess (ELO 2200)
You'll be joining an incredible team:
1. Ben, Software Engineer Intern, https://www.linkedin.com/in/ben-hung/, beat out dozens of seniors on our coding challenges while holding a 3.9+ GPA in Berkeley EECS and 2200 in Rapid chess
It seems the core of the game is correctly understanding the victory function (time to complete refinement of all resources) and then applying bottleneck theory to correctly optimize each stage.
foam.ai | Founding Engineer | $180-$300k | 0.25%-2% | SF (onsite) | Full Time
foam is building a new observability stack that captures, traces, and links everything, purpose-built for AI. We use this foundation to feed complete context into AI to automatically identify and root-cause breakages across entire systems.
* We’re a five person team. Perla (CEO) was the 5th engineer in the consumer facing team at Affirm (pre-IPO).
* We've raised $9.9M in seed funding from Khosla Ventures, Max Levchin, The House Fund, and South Park Commons
* We’ve got active betas at Orb, Braintrust, Plaid & TogetherAI.
We’re looking for engineers who want to take point and work hand in hand with Perla on one or more of our core systems:
1. The Telemetry Engine - Think of this as a major extension of OTEL custom built to capture signals that have historically been inaccessible to humans.
2. The Root Cause Analysis Engine (aka Solver) - This is currently an agent swarm, but evolves every week in search of better and better root cause analysis.
If putting an end to awful on-calls and building a system that can deliver self-healing software is exciting to you, send an email to luke -at- foam dot ai with your resume and mention you came from HN.
foam.ai | Full Stack Founding Engineer | SF (onsite) | Full Time
At foam we’re building the first telemetry and production system designed for LLMs, giving them complete visibility into and control over running software. Today we’re making money by fixing bugs as they happen, tomorrow we’re the data layer for all production observability tooling.
* We’re a six person team. Perla (CEO) was the 5th engineer in the consumer facing team at Affirm (pre-IPO). Shawn Krisman (CTO) was an early engineer at Yelp, Reddit, Patreon and architect at Affirm (pre-IPO).
* We've raised $8.4M in seed funding from Khosla Ventures, Max Levchin, The House Fund, and South Park Commons
* We’ve got active betas at Perplexity, Orb, Braintrust, Plaid & TogetherAI.
If being #7 is exciting; reach out: send an email to luke -at- foam dot ai with your resume and mention you came from HN
> Farm animals grown with their brains shut off, used as compute substrate for biological neural networks, while their biological functions are controlled remotely.
I’m sorry, you were working on what? Where does one learn more about this concept?
One builds the tools to run the experiments to discover the rules.
The closest are FinalSpark and CorticalLabs, but they both are only using in vitro neurons as the computational substrate.
Neuralink et al. are working in vivo, but they are only doing output and don't have any plans to do input, let alone to actively disrupt normal neural activity and take control of bodily processes.
If you're very interested feel free to drop me a line.
I am one of the co-founder of FinalSpark, I believe biocomputing implies to master in vitro learning, and once you get there it may open new perspectives.
Short term in vitro learning's been shown in the original CorticalLabs paper (20 min), but the sheer difficulty of keeping neurons alive, working, unstressed, fed and oxygenated in vitro while highly active for more than an hour at a time means that no one's been able to show how that short term learning transfers to long term memory.
Using an organism that already provides all the life support for the neurons means that you can have test runs that last hours to weeks with continuous stimulation.
Your tech is easily modifiable so it can do that since it's based on the Intan RHS chips. Happy to have a chat about it since you're one of the two companies that could potentially pull this off right now.
From first principles I'd expect the functional limit not to be one of power generation but of energy transfer. The limit probably manifests as wiring and motors overheating in a full electric drive (I don't think any subs are...) or as sound, heat and fatigue in reduction gears and gear shafts.
Location: San Francisco, CA, USA
Remote: Yes, but in-person preferred
Willing to relocate: No
Technologies: Python, Ruby, React, Salesforce, TypeScript
Résumé/CV: https://docs.google.com/document/d/1ux2NKFfinMR_n5d41uge0O2FstDSc-oI/edit?usp=sharing&ouid=110631342450297019344&rtpof=true&sd=true
Email: luke+hn@lukemercado.com
I'd be particularly interested about head of eng roles at startups or director of eng at a scale up. After the co-founding experience I'm also curious about moving into product management or sales in some capacity. For the right company I'd do some software engineering.
reply