manmal

14 ms·

301.
by manmal 9mo ago
Same trip, w/ the kids, took 10h with an Ioniq and 6h with a diesel.
302.
by manmal 9mo ago
Ok wow, that’s enough HN for today.
303.
by manmal 9mo ago
There's a problem with i18n on the landing page, set my browser to German I see things like "home.alternative_title". Tbh I'm not sure such a site needs i18n at all, Claude was a bit overzealous there ;)
304.
by manmal 9mo ago
Make a pull request.
305.
by manmal 9mo ago
No, most of them actually flicker occasionally.
306.
by manmal 9mo ago
Anyone can build _an_ agent. A good one takes a talented engineer. That’s because TUI rendering is tough (hello, flicker!) and extensibility must be done right lest it‘s useless. Eg Mario Zechner (badlogic) hit it out of the park with his i
307.
by manmal 9mo ago
Is it? Recent new models tend to need fewer tokens to achieve the same outcome. The days of ultrathink are coming to an end, Opus is well usable without it.
308.
by manmal 9mo ago
20% of payroll would cover all that, disregarding marketing which a project like tailwind doesn’t need IMO.
309.
by manmal 9mo ago
I can’t see a good reason real estate wouldn’t go the way car dealerships are going.
310.
by manmal 9mo ago
Which business is 100% not at risk in the next 10 years?
311.
by manmal 9mo ago
Oh, only Linux? /s
312.
by manmal 9mo ago
What costs could tailwind the OSS project have besides payroll and SaaS subs of max 20% the payroll?
313.
by manmal 10mo ago
Neither have I, personally, but I’ve seen reports this can happen on very hard problems, where the goal just cannot be reached from a local optimum. Getting unstuck by trying something new is something a watchdog agent could prompt it.
314.
by manmal 10mo ago
Getting in a loop does still happen, yes. If you run codex in tmux and let another agent just occasionally check on progress, it can be prevented. That’s not even expensive - checking every 30 minutes suffices. The watchdog agent can then
315.
by manmal 10mo ago
I’d bet that an agent would never be able to act the same on an email as I would. It just lacks my world view. This begs the question, would it really make sense for it to write emails on my behalf? It will certainly “close the loop” one wa
316.
by manmal 10mo ago
Some of the world is running on emails and excel sheets. That’s doable for an agent already, if you’re willing to let it loose. Problem is, how do you get all your values and unknown knowns into its context?
317.
by manmal 10mo ago
Off/nearshoring regularly produces worse code. I’ve seen it first hand.
318.
by manmal 10mo ago
Just let it test in different containers? That’s not the hard part IMO.
319.
by manmal 10mo ago
As long as you give it deterministic goals / test criteria (compiles, lints, tests, E2E tests, achieve 100% parity with existing solution etc) it will brute force its way to a solution. Codex will work for hours/days, even weeks s
320.
by manmal 10mo ago
I agree, codex is great at reviewing as well. I think that’s because code is the ideal description of what we want to achieve, and codex is good (only) when it knows what must be achieved, as verbosely as possible. Currently I don’t let GLM
321.
by manmal 10mo ago
Is that new though? Software has been hype and marketing driven forever.
322.
by manmal 10mo ago
IMO codex produces working code slowly, while Opus produces superficially working code quickly. I like using Opus to drive codex sessions and checking its output. Clawdbot is really good at that but a long running Claude Code session with c
323.
by manmal 10mo ago
Both clawdbot and pi have improved and expanded functionality a lot during the last 3 weeks, maybe worth another look? What you have described sounds a lot like the experience I’m having.
324.
Clawdbot – Personal AI Assistant(github.com)
2 points by manmal 10mo ago|0 comments
325.
by manmal 10mo ago
Setup is still rough around the edges (use an agent to set it up), but clawdbot (prev clawdis) from Peter Steinberger works phenomenally well for agent orchestration and personal assistance. The community for clawd is exploding right now, a
326.
by manmal 10mo ago
They have ion thrusters that prevent them from losing altitude as long as they are operational.
327.
by manmal 10mo ago
Is there the possibility to hook in OCR for text blocks flattened into an image, maybe with some callback? That’s my biggest gripe with dealing with PDFs.
328.
by manmal 10mo ago
GPT-5.1 was based on over 15 months old data IIRC, and it wasn’t that bad. Adding new layers isn’t that expensive.
329.
by manmal 10mo ago
They are retrained every 12-24 months and constantly getting new/updated reinforcement learning layers. New concepts are not the problem. The problem is outdated information in the training data, like only crappy old Postgres syntax in
330.
by manmal 10mo ago
What’s bad about them? We make things baby-safe and easy to grasp and discover for LLMs. Understandability and modularity will improve.

More ›