Update: More agents via agent factory and Fed thinking

2026-08-31 – blog

The biggest catalyst for the Fed has just passed and I’m a bit bored, so I’ve started on my plan to add more agents. I’m building an agent factory, where one agent helps to construct the prompts for new agents. Kinda funny hahaha. I think this process is also automatable as building the Fed agent took quite a long time to figure out, but now I have a better idea of the process and what I want to see in my agentic research, so I think I will be able to make my agent factory a reality.

I don’t think the new natgas and oil agents are up to the same quality as the fed agent just yet, I think the prompt still needs further refining and same with my agent builder. However, I do want to start following more markets while I wait for September.

As for September and the Fed, I think it boils down to whether the Fed will follow the current data or whether it will follow credibility. Speaking with friends, I’ve broadened my view a bit as I hadn’t totally thought that credibility was a very strong argument for a hike this September. I’m not sure what to do. My agent says to side with data and likes the current odds. Had I not spoken with my friends, I would have liked these odds more. I should be happy though that there is great uncertainty because with great uncertainty means great opportunity.

I still think there’s a reasonable case for the Fed waiting another meeting, but I do think the hike is inevitable on the grounds of credibility and stagflation. If not September, then there will have to be further benign or disinflationary data in Octover and beyond to justify another hold and while I don’t have expertise in predicting inflation I’d be doubtful of disinflationary data. I think the best play is probably to bet on hike by end of this year and hike by end of next year as the Fed must maintain its credibility in the long run.

I suspect if September is a hold, the hike by end of year will dip which I think is a good opportunity to add to a position. But I think now is the best time to take a position.

As for the other agents, I thought it was interesting that natgas is at record storage levels since 2016 and yet there’s coin flip odds on whether the max natgas price hits $6 (representing a 100% increase) by end of this year. The agent also mentioned el nino, but I heard a good argument against that previously. I will have to look into this more I think there is something there.

Anyways, I’ll leave you with the actionable ideas that my agent came up with. As always use your own discretion and do your own research, this isn’t financial advice.

Fed agent

  1. KXFEDDECISION-26SEP-H0 Odds: yes 0.43/0.44 – Sept hold undervalued near 50/50; 54% hike priced against benign core (CPI2.4%, PCE0.2% MoM) Warsh sees through.
  2. KXCPICOREYOY-26SEP-T2.6 Odds: NO 0.72/0.87 – Core CPI at 2.5% July decelerating; ≤2.6% Sept near-certain (>72%), so YES at 28c too rich.
  3. KXIRANCRUDE-26SEP10-T2.6 Odds: NO 0.63/0.70 – Iran ≥2.6M bpd overpriced; Larak re-strikes + maximum sanctions cap output ~1.8-2.4M.
  4. FEDHIKE-26DEC31 Odds: NO 0.28/0.29 – 71% any-hike-EOY vs falling WTI (to ~$75-78 by Nov) + cooling core = inconsistent, likely overshoot.

Oil agent

  1. KXFEDDECISION-26SEP-H0 – Odds: 0.41/0.42 – Market prices 57-58% Sept hike while core CPI stays ~0.2% benign, exactly Warsh’s “look-through” case.

  2. KXCPICORE-26AUG-T0.2 – Odds: 0.32/0.37 – Core CPI MoM ~0.2% is the modal/benign path supporting the no-hike; cheap confirmation of idea 1.

  3. KXWTIMAX-26DEC31-T130 – Odds: 0.17/0.18 – No-deal (8-9%) + heating +87.5% y/y + distillate record-low pre-winter make a $130 year-end spike under-priced.

  4. KXIRANCRUDE-26SEP10-T2.4 – Odds: NO 0.29/0.37 – Market prices 63-71% for Iran ≥2.4M bpd under blockade reality ~1.8-2.4M; overpriced ~20-30pp.

  5. KXWTIMIN-26DEC31-T65 – Odds: NO 0.56/0.61 – Lower confidence; a sub-$65 year-end floor needs a Warsh hike/demand destruction or a deal (92% no), neither priced.

Natgas agent

  1. KXNGASMAX-26DEC31-P5.00 (NO/against) — YES 0.51/0.60 → buy NO at 0.40/0.49. Winter-cold tail over-bid: 51% for >$5 spot vs fair ~20-30% given record 3,985 Bcf storage (highest since 2016) + ~+4°C Super El Nino mild N-tier winter + quiet hurricane season removing the main $5+ path.

  2. KXNGASMAX-26DEC31-P4.50 (NO/against) — YES 0.64/0.71 → buy NO at 0.29/0.36. EIA’s own STEO has HH ≤$3.14 for 4Q26 and $3.03 5-mo avg, so $4.50+ spot is a tail requiring a Gulf-LNG event that the ~50%-no-major hurricane pricing says is unlikely.

  3. KXNGASMAX-26DEC31-P6.00 (NO/against) — YES 0.21/0.28 → buy NO at 0.72/0.79. $6+ needs cold AND an LNG/hurricane squeeze coincidence; quiet Atlantic makes it very remote, offering the widest per-unit edge (I consider this lower-probability but cleanest risk/reward).

  4. KXNGASMIN-26DEC31-N2.40 (NO/against) — YES 0.28/0.30 → buy NO at 0.70/0.72. LOWER-CONFIDENCE take: though the glut argues for a low spot, EBW’s medium-term relief (South Central deficits + rising Gulf LNG feedgas) plus a Freeport-restart feedgas surge should floor HH ~$2.62-2.80, making sub-$2.40 unlikely.

— ✦ —