HNHacker News
TopNewBestAskShowJobs

Bnjoroge

322 karma · joined April 13, 2020

submissionscomments
Bnjoroge··on Show HN: machine0 – Persistent NixOS VMs You Control from the CLI
Very cool. I have something similar set up for my homelab where I launch nixos incus containers. Been great for offloading stuff from the my agents
Bnjoroge··on Kimi K2.7-Code: open-source coding model with better token efficiency
Openrouter has them as significantly more expensive.
Bnjoroge··on MiMo Code is now released and open-source
I mean you can if you have the compute. They’re all open source
Bnjoroge··on MiMo Code is now released and open-source
Indeed. I did the math and arrived at the same conclusion. They don’t really subsidize their token plans. Maybe because their api pricing is already dirt cheap
Bnjoroge··on Kimi K2.7-Code: open-source coding model with better token efficiency
Output tokens are almost 5x more expensive than mimov2.5 pro/dsv4pro. I’m curious to see if Kimik2.7 is that much better. Feels like kimi are positioning themselves as the premium open source models
Bnjoroge··on Kimi K2.7-Code: open-source coding model with better token efficiency
huh? They all work great in omp/opencode unless you mean their own native clis like kimi code
Bnjoroge··on Kimi K2.7-Code: open-source coding model with better token efficiency
It depends on what your task is and how precise your prompts are. Planning with fable or 4.8 and laying out the plan in step by step process and coding with mimo v2.5 pro or dsv4pro or qwen 3.7 max and doing a final review with 5.5 has worked really well for me for infra stuff.
Bnjoroge··on Kimi K2.7-Code: open-source coding model with better token efficiency
I personally dont put any weight to DeepSWE. Other than 5.5 being directionally the best model, it gets the others pretty wrong in my experience. FrontierCode from cognition looks interesting
Bnjoroge··on Kimi K2.7-Code: open-source coding model with better token efficiency
Yup. I’m hoping this variant fixes these issues.
Bnjoroge··on Kimi K2.7-Code: open-source coding model with better token efficiency
This is generally been my experience as well, but i think the main reason for claude code being better at understanding intent is their massive system prompt.
Bnjoroge··on Kimi K2.7-Code: open-source coding model with better token efficiency
They are far from unusable. They aork great for 80-90% of a typical full stack dev. Alot less useful for more noche stuff
Bnjoroge··on Show HN: HelixDB – A graph database built on object storage
it does, think I misunderstood your value prop. best of luck -definitely a real usecase.
Bnjoroge··on Show HN: Command Center, the AI coding env for people who care about quality
love to see jj support, but if I cant use my subs, it's a deal-breaker for me.
Bnjoroge··on Show HN: Command Center, the AI coding env for people who care about quality
plannotator's made me actually want to read my agents' plans. Great product. I've been liking the review as well.
Bnjoroge··on Show HN: Nucleus – A security-hardened, Nix-native container runtime
pretty cool but docker support is a no-brainer. not having it is a deal-breaker
Bnjoroge··on Show HN: HelixDB – A graph database built on object storage
congrats! how does this compare to turbopuffer, surreal or other multi-model ones built on object storage or not
Bnjoroge··on Show HN: Lowfat – pluggable CLI filter that saved 91.8% of my LLM tokens
Yes. That’s possible in addition to using your actual subscription. I’ve been using it via cliproxy for all harnesses and even my own code review agent hooked up to github apps. Not banned yet but I also dont do crazy stuff with openclaw or hermes
Bnjoroge··on Microsoft's open source tools were hacked to steal passwords of AI developers
Docker isn’t a serious sandboxing strategy
Bnjoroge··on Failing grades soar with AI usage, dwindling math skills in Berkeley CS classes
an even lazier comment
Bnjoroge··on Show HN: Lowfat – pluggable CLI filter that saved 91.8% of my LLM tokens
definitely doesnt is a strong word. it technically is possible, but you might get banned
Bnjoroge··on Show HN: Boxes.dev: ditch localhost; run Claude Code and Codex in the cloud
Works well for very well defined task. If you have a really big feature like a front end migration, you can use /plan, and /goal which i think is in most harnesses. You can also use other tools that allow your agent to interact with other terminals(I use an ADE called orca) that has an orca skill where an agent can spin up different sessions(different from subtasks because they share the context and you can chose the harness/model unlike sub agents). Can also read from the terminal, use your browser or computer and task screenshots and after prepare a report or something.
Bnjoroge··on Failing grades soar with AI usage, dwindling math skills in Berkeley CS classes
Lmao
Bnjoroge··on Failing grades soar with AI usage, dwindling math skills in Berkeley CS classes
Or your lazy excuse is an even more convenient thing to blame
Bnjoroge··on Show HN: Boxes.dev: ditch localhost; run Claude Code and Codex in the cloud
What are “box-hours”? Regular hours just running in boxes? Do I get charged the same when 1)the agent is doing some external thing say web search that takes a while, and 2) when the agent isnt running(say waiting for my input)?
Bnjoroge··on Self-hosted dev sandboxes with preview URLs (Docker, Go, no K8s)
This is heavily vibecoded, and probably the 100th iteration of sandboxes. In any case, docker isn’t a serious isolation boundary for agents.
Bnjoroge··on Meta workers can opt out of being tracked at work up to 30 min
No it’s not unless you’ve never interacted with people in other industries. It’s just as bad there, unless you have specific reasons for tech
Bnjoroge··on Meta workers can opt out of being tracked at work up to 30 min
That makes no sense. They didnt provide any specific reasons they dont like the tech industry. All of their reasons can be applied to just about any industry lmao. The core issue is and always will be capitalism
Bnjoroge··on Meta workers can opt out of being tracked at work up to 30 min
None of that is specific to this industry. It’s far worse in others. Grass always looks greener elsewhere
Bnjoroge··on Claude Opus 4.8
Agree on both counts. Mimo seemed to have reduced their prices significantly so if it’s comparable to deepseek v4 pro, it’s a much better value
Bnjoroge··on Using Tailscale with an OrbStack VM on macOS
I use Tailscale with orbstack machines which allows me to run two tailnets at once($dayjob and homelab). Both great products. https://bnjoroge.com/posts/connecting-to-multiple-tailscale-...
← PreviousPage 4 of 9Next →